Exosomal DNMT1 mediates cisplatin resistance in ovarian cancer

Ovarian cancer is the most common malignancy in women. Owing to late syndromic presentation and lack of efficient early detection, most cases are diagnosed at advanced stages. Surgery and platinum‐based chemotherapy are still the standard care currently. However, resistance invoked often compromises the clinical value of the latter. Expression of DNA methyltransferase 1 (DNMT1) was analysed by gene array. Protein was determined by immunoblotting. Exosome was isolated with commercial kit. Cell proliferation was measured by CCK8 method. Annexin V‐PI double staining was performed for apoptosis evaluation. Xenograft model was established and administrated with exosome. Tumour growth and overall survival were monitored. We demonstrated the upregulation of DNMT1 in both tumour and derived cell line. DNMT1 transcripts were highly enriched in exosomes from conditioned medium of ovarian cells. Co‐incubation with exosomes stimulated endogenous expression and rendered host cell the resistance to cytotoxicity of cisplatin. In vivo administration of DNMT1‐containing exosomes exacerbated xenograft progression and reduced overall survival significantly. Moreover, treat- ment with exosome inhibitor GW4869 almost completely restored sensitivity in resistant cells. Our data elucidated an unappreciated mechanism of exosomal DNMT1 in cisplatin resistance in ovarian cancer, also indicating the potential of the combination of exosome inhibitor with cisplatin in resistant patients.

1| INTRODUCTION
Ovarian cancer is one of the most common malignancy in female reproductive system, incidence of which is only lower than cervical cancer and uterine cancer.1 However, ovarian epithelial tumour is the first cause of death in all gynaecological tumours and imposes severe threat on women’s health worldwide. Clinically, ovarian cancer is usu- ally diagnosed at late stages due to lack of specific and sensitive early diagnosis methods.2 The patients with advanced disease are not ame- nable to surgical removal because of multiple distal metastasis. Simi- larly, treatment effect of radiotherapy is very limited for this disease.3 Currently, preoperative systematic neoadjuvant chemother- apy is the standard procedure especially for advanced genital cell cancer, which could create a permissive condition for surgical resection right after tumour shrinkage.4 Platinum‐based chemicals are the first widely used drugs in bench and still the mainstay agents for ovarian chemotherapy nowadays.5 Platinum analogs have broad spectrum of antitumour activity against a variety of malignancies including gynaecological cancers, germ cell tumours, head and neck cancer, thoracic cancers, and bladder cancer.6 Despite enormous therapeutic potency, the resistance invoked seriously compromises the clinical outcome of cisplatin.7 Therefore, elucidation of underlying mechanism is crucial to conquer the resistance. However, previous investigations demonstrated complicated sources for emergence of cisplatin resis- tance in different human cancers, and most of them are associated with de novo persistent genomic alterations such as mutations and copy number variation.

In addition to genetic factors, mounting evidences have also highlighted that epigenetic changes contribute to drug resistance. Aberrant genomic DNA methylation pattern and histone modifications subtly modulate gene expression, which are tightly involved in tumour initiation, progress, recurrence, resistance, etc.9 There are three major DNA methyltransferases (DNMTs) in mammals: DNMT1, 3A, and 3B. DNMT1 is the one responsible for maintenance of genome‐wide methylation during DNA replication and repair,10 while 3A and 3B catalyse de novo methylation and established newly acquired epigenetic markers.11 Although many investigations demonstrated that gene variants of DNMTs are linked to the predisposition of a variety of human cancer, the relationship between DNMT1 expression and chemotherapeutic resistance has not been explored.Exosomes are microvesicles secreted by host cells via fusion of multivesicular bodies with the plasma membrane.12 Intracellular materials are strictly sorted and selected for exosome packaging.13 Exosomal composition and signature vary greatly among each other in circulation and tissue microenvironments and are completely depen- dent on the cell origin. Accumulating evidences suggest that the exosome plays crucial physiological roles in processes such as coagula- tion, intercellular communication, and waste management.14 Exosomes were also reported to be involved in tumour initiation, progress, recurrence, and therapeutic resistance via transmission of microRNA, lncRNA, mRNA, and protein factors.15 However, exosome‐mediated modulation of epigenetic factors, especially DNMT1 in ovarian cancer, has not been thoroughly investigated. Here we sought to explore the relation between DNMT1 expression and cisplatin and elucidate the involvement of exosome in this process.

2| MATERIALS AND METHODS
Human ovarian cancer cell SKOV3 and normal endometrial stromal cell line ESC were obtained from and authenticated by the American Type Culture Collection. Cells were regularly inspected for mycoplasma contaminations. Both cells were cultured in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% foetal bovine serum (FBS) and 1% penicillin‐streptomycin. Cells were maintained in humidified 37°C incubator supplied with 5% CO2. Exponentially growing cells were harvested for all experiments in this study.Cisplatin was dissolved in distilled water at 10 μg/mL and stored in aliquots at −20°C until use. Final concentrations between 0.5 and 10 μg/mL of cisplatin were obtained by appropriate dilutions of the stock compound with the defined medium. In 96‐well plate, 2 × 104 cells in 100 μL of DMEM were seeded the day before treatment. Cells were drugged with indicated concentration of cisplatin. The cell viabil- ity was evaluated by MTT assay. MTT (20 μL) solution (0.5 mg/mL) was added to each well and incubated for 4 hours. The medium was aspi- rated and replaced with 150 μL DMSO. After 15 minutes of agitation, the absorbance at 490 nm was measured by Versamax microplate reader (Bio‐Rad, Hercules, CA, USA).The medium from SKOV3 and ESC cells were collected and centri- fuged at 3000 rpm for 10 minutes. The supernatant was preserved as conditioned medium (CM) for the following assays. Exosomes were isolated from conditioned medium using the ExoQuick‐TC Exosome Precipitation Solution (System Biosciences, Mountain View, CA, USA).

Briefly, the equal volume of Precipitation Solution was added to the conditioned medium and incubated at 4°C overnight. The samples were serially centrifuged by 1500 rpm for 30 minutes and then 3000 rpm for 5 minutes at 4°C. The exosome pellet wasresuspended in 250 to 500 μL of appropriate cell medium. For exosome depletion, CM was ultracentrifuged overnight. For exosome transfer assays, an equal volume of exosomes was added to cells for24 to 48 hours of consecutive culture. RNA species included in exosome were extracted after 48 hours of culture using TRIzol method. Exosome was lysed in RIPA buffer in accordance with standard proto- col for protein preparation.The primers used in this study are listed as follows: DNMT1 Forward: 5′‐ GGACGACGGGAAGACCTA ‐3′.Reverse: 5′‐ CTGGAGTGGACTTGTGGG ‐3′;GAPDH Forward: 5′‐CCATCACCATCTTCCAGGAG‐3′. Reverse: 5′‐CCTGCTTCACCACCTTCTTG‐3’;The total RNA was extracted with Trizol reagent. The purity and integrity were checked before further use. The first strand cDNA was synthesized with PrimeScript RT reagent kit (TaKaRa Bio, China). The RT‐PCR was conducted with SYBR Green Master kit (Promega, Madison, WI, USA) following the manufacturer’s instruction. The bands were visualized with EB staining and analysed with ImageJ soft- ware. The relative expression was calculated and normalized to GAPDH. The results are representative of at least three independent experiments.Four‐week‐old female BALB/c‐nude mice were purchased from Beijing Vital River Laboratory Animal Co., Ltd, China. All animals were housed in a pathogen‐free environment, and experimental protocols were approved by the Committee of Animal Care and Use of Cangzhou Center Hospital.Exponentially growing cells were harvested and resuspended in sterile PBS. Equal numbers (2 × 106) of each cell were inoculated subcutaneously into the right flanks of immunodeficiency mice.

Tumour growth was monitored twice a week using digital calliper according to the formula: TV (mm3) = length × width2 × 0.5. Mice were sacrificed at the time indicated, and actual size were determined.Appropriate number of cells was seeded in 96‐well plate subjected to control culture or treatment indicated. Of the CCK‐8, 10 μL solution was added into each well for 1 to 4 hours of incubation. Absorbance at 450 nm was measured by microplate reader.The indicated cells were washed and lysed completely in RIPA lysis buffer on ice for 30 minutes. The lysate was centrifuged to remove cell debris. The total protein was quantified by Coomassie brilliant blue assay. Approximately 15 μg of protein was electrophoresed on 10% SDS‐PAGE gel and transferred to PVDF membranes at 4°C. The mem- branes were blocked with 5% milk dissolved in Tris‐buffered saline plus 0.1% Tween‐20 (TBST) for 1.5 hours and blotted with antibodies against DNMT1, CD9, CD63, and β‐actin overnight at 4°C. Themembranes were washed rigorously with TBST at room temperature and incubated with secondary antirabbit IgG‐HRP (1:5000) for 2 hours at room temperature. The membranes were washed 3 × 6 minutes with TBST at room temperature and visualized by enhanced chemilu- minescence reagent according to the manufacturer’s instruction. The intensity of the individual bands was quantified by densitometry (Bio‐Rad, Hercules, CA, USA) and normalized to the corresponding input control (β‐actin).Targeted quantitative RT‐PCR arrays were used to measure expres- sion of DNMT1. PCR primer pairs were designed with Primer 3 (http://primer3.sourceforge.net/) software. Gene array was constructed by spotting and air‐drying primer (100 nmol/5 μL) into each wells of a Lightcycler 480 Multi‐well Plate 96 (Roche, Penzberg, Upper Bavaria, Germany). Of the reaction, 20 μL cocktail containing cDNA from 5 μg RNA template and SYBR green master mix were added to each well. PCR reactions were conducted in Roche Lightcycler 480 System. Relative gene expression was calculated by ΔΔCt method and normalized to β‐actin.Apoptosis was determined with Annexin V‐PI method. Briefly, the cells were digested with 0.25% trypsin and resuspended in binding buffer to a density of 1 × 106 cells/mL. Add 5 μL of Annexin V‐fluorescein iso- thiocyanate and 5 μL of propidium iodide (Sigma‐Aldrich; the cells were incubated in the dark at room temperature for 15 minutes). Cell death was measured by flow cytometer (FACSCalibur; BD Biosciences, Franklin Lakes, NJ, USA). Data were analysed using CellQuest Pro soft- ware, version 6 (BD Biosciences, Franklin Lakes, NJ, USA).Data from three independent experiments were subjected to variance analysis using SPSS 19.0 software, and all the results were presented as mean ± standard deviation (SD). The statistical significances between data sets were expressed as p values, and p < 0.05 was considered statistically different.

3| RESULTS
Accumulating evidence suggested that aberrant methylation pattern is intimately associated with ovarian tumourigenesis.16 DNA methyl- transferase catalyses methyl group transfer to cytosine nucleotides of genomic DNA and consists of three family members: DNMT1, DNMT3A, and DNMT3B in mammals. DNMT1 is the major enzyme responsible for maintaining methylation memory during DNA replica- tion and shows a preference for hemi‐methylated DNA, while DNMT3A and 3B predominately catalyse de novo methylation, which is crucial for development. Gene variants were frequently identified as driving force in a variety of human cancer. However, the potentialassociation of DNMT1 expression with ovarian tumourigenesis has not been investigated. Here we first characterized the relative expres- sion of DNMT1 in clinical ovarian tumour samples.As shown in Figure 1, we first constructed targeted quantitative RT‐PCR array for measurement of the relative expression of DNMT1 in tissue samples. We employed GAPDH as internal reference for quantitation purpose. In total, we have collected 30 human ovarian tissue samples with paired adjacent nontumour tissues. We have presented our data as heat map for straightforward comparison. Remarkable upregulation of DNMT1 in ovarian samples was observed (Figure 1A). The bar plot showed significant difference of DNMT1 between ovarian cancer and normal control, which was approximately 5‐fold higher in tumour (Figure 1B). The protein level of DNMT1 was also determined by immunoblotting, and representative image was showed in Figure 1C.

All bands were scanned by densitometry, and relative protein content was calculated. Bar plot showed that average DNMT1 protein in tumour was 4 times higher than normal, which is consistent with the change at transcript level. We further verified this phenomenon in cultured cell lines in vitro. For this purpose, we chose a well‐recognized ovarian cancer SKOV3 and normal endometrial stromal cell ESC. mRNA and protein of DNMT1 was 4.5 and 6 times higher in SKOV3 than in ESC, respectively. Our data unambiguously demonstrated overexpression of DNMT1 in ovarian cancer.Exosomes are microvesicles of 70 to 120 nm that originate from fusion of multivesicular bodies with the plasma membrane. Exosomes are increasingly recognized in mediating long‐range and short‐range inter- cellular communication and reconstructing tumour microenvironment, thus playing an important role in tumour initiation, progression, recur- rence, and resistance. Until now, the epigenetic transfer mediated by exosome in cancer has not been investigated. Here we sought to clarify whether exosome was involved in DNMT1 regulation in ovarian cancer.First, we isolated exosomes from conditioned medium of SKOV3 and ESC cell culture. Ultraspeed centrifugation yielded similar amount of exosomes from both SKOV3 and ESC cell lines. Distribution of exosome was analysed by scanning electron microscope, and those in the 90‐ to 120‐nm range were selected for further characterization (Figure 2A). Reliability of our isolation procedure was critically evalu- ated by immunoblotting with exosome‐specific marker CD9 and CD63 (Figure 2B). Total RNA was extracted from exosomes by regular Trizol method and subjected to array (Figure 2C) and PCR analyses (Figure 2 D).

Relative content of DNMT1 was significantly higher in exosome from SKOV3 cell (p < 0.01), which highly likely suggested that DNMT1 mRNA was specifically packaged into exosomal cargo for epigenetic regulatory signalling transmission, and this event might be functionally indispensable for tumour growth.Next, we set out to elucidate the pathological significance of exosomal DNMT1 transcripts in ovarian cancer. SKOV3 cells were brieflyexposed to highly concentrated exogenous exosome, and then endog- enous mRNA of DNMT1 was measured by RT‐PCR. A doubled increase was observed upon treatment, and this appeared to be attrib- uted to the internalization of exosomal DNMT1 mRNA despite that we could not distinguish the origin based on our current data (Figure 3A). Elevation of DNMT1 would greatly shift epigenetic landscape; there- fore, we challenged SKOV3 cells with first‐line chemotherapy agent cisplatin to explore the potential impact of exosomal DNMT1 on drug resistance. Examination of cell proliferation with CCK8 assay showed that exosome markedly increased survival under 3μM cisplatin treat- ment (Figure 3B). Kill curve drawn from MTT results demonstrated sig- nificant resistance in SKOV3 with exosome in comparison with SKOV3 control (Figure 3C). In untreated SKOV3 cells, incubation with 3μM cisplatin elicited massive apoptosis, while exosome‐exposed cell showed significant resistance to this induction (Figure 3D). From our data, it can be proposed that intracellular DNMT1 expression could be elevated by exogenous exosome, which was enriched with DNMT1transcripts, and then contributed to cisplatin resistance in human ovarian cancer.Our previous data demonstrated that endogenous DNMT1 transcripts were highly specifically packaged into exosome, which were released to extracellular milieu and in turn promoted DNMT1 expression in host cells. The resultant high expression of DNMT1 thus contributed to cisplatin resistance in our ovarian cancer cell through epigenetic alter- ations.

Then, we further consolidated this possible mechanism in vivo. To this end, we established ovarian tumour xenografts by inoculat-ing SKOV3 single cell suspension subcutaneously. The treatment group mice were administrated with exosome isolated from SKOV3‐ conditioned medium by tail vein injection. The xenografts progressed rapidly when complementing with exosome intravenously incomparison with the control (Figure 4B). Average tumour size in the control group was 3200 μm3, while it was enlarged to 7500 μm3 in the exosome group. Concomitantly, the predicted lifespan expressed as overall survival was significantly shortened in exosome‐adminis- trated mice (Figure 4A), dropping from 75 to 51 days after inoculation. Although we have not determined the response to chemotherapy in this setting, our results clearly showed that DNMT1‐containing exosome dramatically accelerated ovarian tumour progression in vivo.Our previous data suggested an unappreciated mechanism of exosomal DNMT1 underlying progression and drug resistance of human ovarian cancer. Then, we attempted to exploit this for potential therapeutic intervention. For this purpose, we utilized GW4869, a spe- cific inhibitor of nSMase2 to inhibit exosome release, to block exosome‐mediated DNMT1 transmission. We first cultured SKOV3 cells with GW4869 in medium, and exosome was collected. The enriched exosome was then in turn applied on SKOV3 again, and its impact on DNMT1 expression was determined by RT‐PCR. Our results showed that GW4869 treatment almost completely abolished the induced increase of DNMT1 in SKOV3 cells (Figure 5A). The resistance to cisplatin in SKOV3, which was imposed by exosomal DNMT1, was remarkably reversed by GW4869 blockade (Figure 5B). On the other hand, spontaneous apoptosis during SKOV3 culture was greatly inhibited by the addition of DNMT1‐containing exosome. However, in our experiment, interference exosome secretion with GW4869 ledto the impairment of its apoptosis suppression (Figure 5C). Similarly, apoptotic response to 3μM cisplatin was recovered to a comparable level with the control, which was dramatically declined in the exosomal DNMT1 treatment group (Figure 5D). Our data clearly showed that blockade of exosome release might relieve the resistance to cisplatin in our setting, and elucidation of this pathway could offer more target for therapeutic exploitation.

4| DISCUSSION
Ovarian cancer is one of the most lethal gynaecological malignan- cies.2 Every year 220 000 women are diagnosed with epithelial ovarian cancer worldwide, and approximately 15 000 cancer‐related deaths are claimed in the USA. Most new cases are at late stage according to the International Federation of Gynecology and Obstetrics stratification and partially due to its late presentation of syndromes. Currently, the standard care remains to be surgical removal, platinum‐based cytotoxic chemotherapy, and neoadjuvant chemotherapeutics, which is a well‐accepted preoperative adminis- tration for conditions when surgical debulking is not feasible at the time of diagnosis. The prior treatments with chemicals elicit tumour shrinkage and provide optimal windows for surgical interventions.With the advancement of our understanding on ovarian tumourigenesis at molecular level, many other therapeutic strategies have been developed for clinical trials. Targeted therapy is exploited to aim at recurrent mutations identified in large‐scale cancer patients with the aid of next generation sequencing technology. For example, poly(ADP‐ribose) polymerase (PARP) inhibitor displays great clinical potential in BRCA mutant patients, which are deficient in effective homologous recombination DNA repair pathway and thus heavily dependent on mismatch repair capacity.17 Targeted inhibition of PARP elicits synthetic lethality with intrinsic defect in DNA damage response. Rapidly proliferative tumour demands massive blood vessel neogenesis for nutrient supply and distal spread. Clinical trials with the monoclonal antivascular endothelial growth factor antibody, bevacizumab, present encouraging primary results.18 In recent years, immune checkpoint inhibitors have greatly shifted the paradigm of cancer treatment.19 Enormous endeavours have been invested in the exploration of the potential clinical efficacy of immune checkpoint inhibitors in ovarian cancer. Relatively limited outcomes have been reported so far, which implicates that more accurate subtyping and biomarker methods are in urgent need for this purpose.20 Platinum‐ containing regimens constitute the first‐line chemotherapy and have been the gold clinical standard for almost 40 years. The options have been revolutionized continually with the combination of other avail- able agents such as paclitaxel and docetaxel plus less toxic carboplatin.21

Despite of its widely appreciated therapeutic efficacy in ovarian tumour, drug resistance appears to be the inevitable endpoint of cisplatin.8 Many investigations attempted to address this issue, and a variety of molecular mechanisms contributing to cisplatin resistance has been proposed. Drug resistance has been demonstrated that could develop along alterations that reduce active intake of cisplatin via partial inactivation of transporter and activation of exclusion.22 MRP2, the well‐characterized multidrug resistance‐associated protein 2 that expels excessive intracellular cisplatin actively, overexpression has been identified in ovarian cancer, thus may be linked to its resistance.23 The sensitivity of cancer cells to the cytotoxic effect of cisplatin is dominated by the presence of a proficient DNA repair apparatus, especially the nucleotide excision repair that which specifically resolves DNA‐cisplatin adduct. Elevated nucleotide excision repair and homologous recombination capacity have been reported in cisplatin‐resistant human cancer.24 The cytotoxicity effect of cisplatin relies on sensing DNA damage and converting to cell lethal signalling. Genetic or epigenetic aberrance impairs downstream pro‐apoptotic signalling and eventually undermines the sensitivity. There are also evidences supporting the compensatory mechanisms that deliver pro‐proliferative and survival cues via molecular circuit in response to the cytotoxicity imposed by platinum treatment. For instance, overexpression of receptor tyrosine kinase ERBB2 has been characterized in ovarian cancer to promote cisplatin resistance via robust growth stimuli.25 Similarly, upregulation of dual‐specificity Y‐phosphorylation‐regulated kinase 1B appears to sustain cisplatin resistance as it favours expression of various antioxidant enzymes.26

In addition to genetic alterations, epigenetic dysregulation is increasingly recognized to function in drug resistance emergence. For example, GPX3 promoter methylation could predict platinum sensitiv- ity in colorectal cancer.27 Lysine‐specific demethylase KDM3A regu- lates ovarian cancer stemness and chemoresistance via inducing the expressions of pluripotent molecules Sox2, Nanog, and anti‐apoptotic Bcl‐2.28 Hypomethylation of CpG sites within the gene body is associ- ated to the low expression of MSX1 and cisplatin resistance in high‐ grade serous epithelial ovarian cancer.29 Promoter hypermethylation was also identified in putative DNA/RNA helicase SLFN11, which was proposed to confer resistance to platinum drugs.30 It was reported that the regulator of G protein signalling 10, RGS10, was suppressed during acquired chemoresistance in ovarian cancer via promoter hypermethylation, which was mediated by both DNMT1 and HDAC.31 Inhibition of DNMT1 and HDAC either by knockdown or by pharma- ceutical agents in this setting restored the sensitivity of cisplatin. Although it is generally recognized as maintenance apparatus for cellu- lar epigenomic landscape, dysregulated expression or activity of DNMT1 is frequently associated with malignancy transformation.32,33 However, the potential link between DNMT1 with drug resistance in ovarian cancer has not been elucidated. In this study, we characterized the upregulation of DNMT1 in ovarian cancer both in vivo and in vitro. We further demonstrated that DNMT1 transcript was specifically packaged into exosome for material transmission, and exosome from condition medium in turn elicits expression increase of DNMT1, which eventually leads to apoptosis suppression and cisplatin resistance. Treatment with the exosome secretion inhibitor GW4869 significantly abolished the resistance in our experiment. In conclusion, our study uncovered an unappreciated mechanism that exosome modulated the overexpression of DNMT1 contributing to cisplatin resistance in ovarian GW4869 cancer.