Role of the Exosome Secretion Machinery in Ovarian Carcinoma: In Vitro and In Vivo Models

Objective . We recently reported on the expression and clinical role of molecules that mediate exosome secretion in high-grade serous carcinoma. In the present study, the biological role of these molecules was analyzed. Methods . OVCAR8 and ES-2 ovarian carcinoma cells were studied using a combination of CRISPR/Cas9 technology and two 3D in vitro models—spheroids emulating eﬀusions and alginate scaﬀolds representing solid lesions. Isolation of exosomes was validated by electron microscopy. TSAP6 , NSMASE2 , RAB27A , and RAB27B mRNA and protein levels were analyzed using qRT-PCR and Western blotting, respectively. Tumor aggressiveness was studied in vitro using scratch assay, invasion assay, and matrix metalloproteinase (MMP) activity assay and in vivo using a mouse model. Results . In OVCAR8 cells, mRNA expression of TSAP6 and RAB27A was signiﬁcantly higher in spheroids compared to scaﬀolds, whereas the opposite was true for NSMASE2 mRNA. In ES-2 cells, TSAP6 and RAB27B mRNA expression was signiﬁcantly higher in spheroids versus scaﬀolds. In addition, nSMase2 and TSAP6 protein expression was signiﬁcantly higher in scaﬀolds compared to spheroids. CRISPR-edited cells with silencing of NSMASE2 , TSAP6 , and RAB27A/B had reduced migration, invasion, and MMP activity. Additionally, knockout (KO) of these molecules resulted in signiﬁcantly diminished exosome secretion. In vivo assay showed that when injected to mice, OVCAR8 RAB27A/B KO cells, as opposed to control OVCAR8 cells, did not form ascites or visible tumor lesions and had reduced MMP expression. Conclusion . The present study provides evidence that diﬀerent models for culturing ovarian carcinoma cells aﬀect the expression of molecules mediating exosome secretion and that these molecules have a tumor-promoting role. Silencing these molecules may consequently have therapeutic relevance in this cancer.


Introduction
Ovarian cancer, the most aggressive gynecologic cancer, was predicted to be diagnosed in 22,530 women and led to 13,980 fatalities in the USA in 2019 [1]. Global figures for 2018 are 295,414 new cases and 184,799 deaths, making this cancer the 8 th most common and 8 th most lethal cancer in women [2]. While novel targeted therapies, such as PARP inhibitors, have extended the survival of patients carrying mutations in BRCA and related DNA repair genes, prognosis remains dismal for the majority of patients, particularly when the disease is diagnosed at advanced stage [3].
Exosomes derived from tumor cells have been shown to have cancer-promoting effects, including increased motility through extracellular matrix (ECM) [13], increased degradation of ECM [14], and invasiveness [15]. Cancer-derived exosomes can also coerce surrounding cells to contribute to their cancer microenvironment: tumor exosomes can transform surrounding fibroblasts into cancer-associated fibroblasts. Exosomes can also cause increased vascular permeability [16], contributing to the metastatic capability of the tumor, as well as condition premetastatic tumor niches at various distant anatomic sites [17,18].
A previous study by our group reported on anatomic site-dependent expression of components of exosomes secretion machinery, i.e. neutral sphingomyelinase2 (nSMase2), Tumor suppressor-activated pathway 6 (TSAP6), Ras-related protein 27A/B (Rab27a/b) and ADPribosylation factor 6 (ARF6) in high-grade serous carcinoma (HGSC), the most common and clinically aggressive histologic type of ovarian carcinoma (OC). TSAP6 and RAB27a levels in effusion specimens were additionally significantly related to overall survival [19].
Another previous study by our group, focusing on Ezrin expression in OC, suggested that tumor growth in a spheroid model closely resembles tumor growth in ascites and pleural effusion, whereas cells cultured on macroporous scaffold are a model for solid tumor growth, thereby creating an inclusive three-dimensional (3D) in vitro model that emulates the different anatomic sites affected by OC [20]. In the present study, we utilized this model together with CRISPR-Cas9 Knockouts (KO) in OC cell lines. Our objective was to demonstrate that NSMASE2, TSAP6, and RAB27A/B-related pathways are essential for exosome secretion in OC cell lines and may consequently constitute a potential therapeutic target in OC.

Cell Culture.
e OVCAR8 and ES-2 OC cell lines were purchased from ATCC (Manassas VA) and cultured at 37°C, 5% CO 2 in their appropriate medium (RPMI + 5% FCS and DMEM + 10% FCS, resp.), supplemented with 1% nonessential amino acids, 1% L-glutamine, 1% BME vitamins, 1% sodium pyruvate and 1% penicillin, streptomycin, and amphotericin. All reagents were purchased from Biological Industries (Beit HaEmek, Israel). OVCAR8 and ES-2 OC spheroids were formed by constant shaking of 400,000 cells per well in 6-well plates for 48 hours on a vertical shaker. Alginate scaffolds with a diameter of 5 mm and thickness of 2 mm were prepared as high guluronic acid (LVG) alginate (NovaMatrix FMC Biopolymers, Drammen, Norway). e macroporous structure was obtained by employing a freezedrying technique: alginate was dissolved in DDW (1.2% w/v solution), homogenized and cross-linked with a solution of calcium gluconate (1.2% w/v) by homogenizer apparatus. e final concentration of cross-linked solution was 1.0% and 0.2% w/v for the polymer and cross-linker, respectively. e cross-linked solution was doled out into 96-well plates (100 μL/well) and then transferred to +4°C for 1 hour, subsequently to −20°C for 24 hours, and then lyophilized. 400,000 cells were seeded on each alginate scaffold and were cultured for 1-2 weeks.

Exosome Extraction and
Quantitation. 4 × 10 6 cells were seeded in 100 mm dishes and incubated overnight with supplemented medium of RPMI 5% FCS. On the next day, the medium was replaced with serum-free medium (EXCELL advanced CHO; Sigma-Aldrich, St. Louis MO) and cells were incubated for further 48 hours. e medium was collected and filtered with Amicon Ultra-15 Centrifugal 0.1 μM PVDF Filter Units (Merck Millipore Ltd). Next, the medium was concentrated with 3000 MWCO vivaspin 20 (Sartorius, Göttingen, Germany). Exosomes were extracted from the supernatant according to the user's manual of ExoQuick-TC (System Biosciences, Mountain View CA). e resulting pellet containing exosomes was resuspended in 100 μl of PBS and quantified for protein concentration by Bradford assay. Exosomes were detected and quantified using the qNano Gold analyzer (IZON Science Ltd., Christchurch, New Zealand).

Electron Microscopy (EM).
Exosomes extracted from all genotypes examined were negatively stained with Sodium phosphotungstate (PTA) and imaged by a transmission electron microscope, JEM-1400 plus, available in the Inter-Departmental Equipment Department at the Ein Karem campus of the Hebrew University.
is procedure was performed by the staff of the department according to standard protocol. OVCAR8 control-derived exosomes were diluted 1 : 1000 in PBS.

qRT-PCR.
RNA was extracted with Tri-Reagent (Sigma-Aldrich) and converted to cDNA with qScript cDNA synthesis kit (Quanta Biosciences, Gaithersburg MD). qPCR was performed with KAPA SYBERFAST Universal qPCR kit (Kapa Biosystems, Wilmington MA) on the CFX Connect Real-Time system (Bio-Rad Laboratories, Hercules CA). e final concentrations of template and primers were determined individually for every assay, calibrated based on a standard curve. Specific primer sequences were designed utilizing the NCBI Primer Blast tool (Table 1). Analysis was performed with BioRad CFX Manager Software. e ΔCT of genes of interest was calculated relative to the RPLP0 housekeeping gene and expressed as 2 -ΔΔCT versus average primary as a control, except for TSAP6 which was analyzed as TSAP6/RPLP0.
When necessary, standard PCR was run using the above equipment and primers with 2G FAST KAPA master mix (Kapa Biosystems). e products were run with electrophoresis on 1.5% agarose gel stained with ethidium bromide, imaged, and then analyzed by ImageJ.

CRISPR-Cas9 Knockouts (KO). Clustered Regularly
Interspaced Short Palindromic Repeats-(CRISPR-) associated 9 (Cas-9) KOs were generated in OVCAR8 cells using the pSpCas9 (BB)-2A-Puro (PX459) plasmid obtained from Addgene (Watertown MA) according to the manufacturer's protocol. Briefly, specific single guide RNA (sgRNA) inserts were designed to target the beginning of the 5′ sequence of every gene (Table 2). e inserts were 5′ phosphorylated and annealed on a ramp between 95°C and 25°C. e vector was digested by the BBSI restriction enzyme. Ligation of the insert to the plasmid was performed with T4 ligase. e plasmids were then transformed to DHα5 E. coli cells. Plasmids were extracted from the DHα5 E. coli cells with GeneJet Plasmid Miniprep (Fermentas Life Sciences) and were sent to the Center for Genomic Technologies at the Hebrew University, Givat Ram, Jerusalem, for sequencing, in order to confirm the insertion of the sgRNA insert. Plasmids were subsequently transfected into OVCAR8 cells with Lipofectamine 2000 (Invitrogen, Carlsbad CA). Selection by Puromycin (A.G. Scientific Inc. San Diego CA) was performed, and the surviving resistant cells were seeded as single cell colonies, resulting in formation of single cell clones of-KO cells, which were then analyzed for KO with PCR or qPCR as described above.

Migration-Scratch
Assay. OVCAR8 cells with NSMASE, TSAP6, and RAB27A/B KOs and OVCAR8 control cells were seeded to confluence in 6-well plates. Each well was scratched twice with a sterile tip. e scratch in NSMASE2 and TSAP6 KO cells was imaged at t � 0 and t � 18 hrs. RAB27A/B KOs were imaged at t � 0 and t � 24 hrs. Wound closure was analyzed by T-scratch software [21].

Matrix Metalloproteinases (MMP) Activity
Assay. KO cells and OVCAR8 control cells (n = 500,000) were incubated overnight in DMEM medium supplemented with 0.1% BSA. e conditioned medium was loaded on 1 mg gelatin/ ml, 10% SDS polyacrylamide gels, and separated by electrophoresis. e gels were incubated overnight at 37°C and then washed in 2.5% triton, developed, dried, and analyzed with the ImageJ software.
2.9. Invasion Assay. Cells of all genotypes (n � 50,000) were seeded on Matrigel-coated filters in Boyden chambers and incubated for 6 hours at 37°C with conditioned medium from the 3T3 fibroblast cell line, used as chemoattractant. Invading cells were stained and quantified under an invert microscope.

In Vivo Tumor Aggressiveness
Assay. OVCAR8, RAB27A-KO, and RAB27B-KO cells (2 × 10 6 ) were injected IP into female SCID mice (n � 12; 5 controls, 4 RAB27A-KO, 3 RAB27B-KO). Mice were monitored for 21 days, until symptoms of OC appeared. e tumors were removed, fixed in formalin, and embedded in paraffin. H&E sections were assessed by an experienced surgical pathologist (BD). e presence of ascites fluid was noted. All experiments were authorized by the Animal Ethics Committee of the Hebrew University of Jerusalem.

Statistical
Analysis. Student's t-test was employed, significance was determined as p < 0.05. High and low mRNA and protein expression was defined relative to expression in control cells in each experiment.

Growth Form in 3D Models Influences Gene and Protein
Expression in OC. Our aim was to develop a 3D model that emulates OC. To this end, we successfully introduced two separate 3D in vitro models utilizing the OVCAR8 and ES-2 OC cell lines-a spheroid model and a scaffold model. Analysis of the genes involved in exosome secretion in OVCAR8 cells showed different NSMASE2, TSAP6, and RAB27A levels as function of the culture method (Figures 1(a)-1(d)). mRNA expression of TSAP6 and RAB27A was higher in spheroids compared to scaffolds (p < 0.01 and p < 0.001, respectively), whereas the opposite was true for NSMASE2 mRNA (p � 0.0001). No change was observed in RAB27B mRNA levels. In ES-2 cells, TSAP6 and RAB27B mRNA expression was higher in spheroids versus scaffolds (p < 0.03, p < 0.02, respectively) ( Supplementary  Figures 1(a) and 1(b)).
nSMase2 and TSAP6 protein expression was higher in scaffolds compared to spheroids (p < 0.006 and p < 0.0004, respectively). e growth form did not affect RAB27A protein expression (Figures 2(a)-2(c), Supplementary Figure S2). RAB27B protein expression was not analyzed due to lack of adequate antibody.

CRISPR Cas9 KO in OC Cell
Lines. In order to evaluate the distinct role of the genes putatively involved in exosome secretion, OVCAR8 cells were subjected to KO with the CRISPR Cas9 protocol for the various genes in this study. KOs were validated by qPCR showing the absence of mRNA and by Western Blot, proving that this gene was silenced and not expressed in this cell line. e percentages of KO achieved and protein levels are detailed in Table 3.

Suppression of Exosome Secretion Pathways Leads to Reduced Migration, MMP Activity, and Invasion in OC Cells.
OVCAR8 cells with KO of any of the four genes, that is, NSMASE2, TSAP6, RAB27A, and RAB27B, migrated significantly less in the scratch assay (Figure 4(a)). After 24 hours, the wounds in OVCAR8 cells with KO mutation of NSMASE2 and TSAP6 were 73% and 55% open, respectively, versus 27% of the OVCAR8 control (p < 0.001 and p < 0.03, resp.).
In the Boyden chamber invasion assay, OVCAR8 RAB27A/B KO cells infiltrated significantly less than control OVCAR8 cells: an average of 32 and 44 cells succeeded in infiltrating, respectively, versus an average of 73 cells in the control (p < 0.0005 and p < 0.05, respectively). A trend was noted for NSMASE2-KO cells where only 52 cells infiltrated (p � 0.087; Figure 4(c)).

Suppression of Exosome Secretion Pathways Leads to Absence of Ascites Fluid and Reduced Tumor Size In Vivo.
Seven weeks following the injection of 2 × 10 6 cells of OVCAR8 control, RAB27A-KO and RAB27B-KO to SCID mice, only the mice injected with OVCAR8 control cells developed ascites, with a volume ranging from 0.75 to 1.5 ml. No ascites ascites was observed in the mice injected with KOs. e entire group of RAB27A-KO and RAB27B-KO mice exhibited no visible tumors by gross examination, whereas tumor nodules were detected in the OVCAR8 control group. Microscopic assessment of the abdominal organs from the 12 mice showed that in RAB27A-KO and RAB27B-KO mice, the majority of organs were tumor-free, with isolated tumor nodules ranging from 0.5 mm to 4 mm and 0.5 to 2 mm, respectively, whereas the OVCAR8 control group had multiple lesions with a diameter of up to 7 mm ( Figure 5, Supplementary Table S1). e average diameter of metastasis, including tumor-negative organs, was 0.59 mm for the RAB27A-KO group (total diameter of all metastases 13.5 mm in 23 organs), 0.35 mm for the RAB27B-KO group (total diameter of all metastases 5.5 mm in 16 organs), and 1.53 mm for the control group (total diameter of all metastases 44.5 mm in 29 organs; p < 0.05).
IHC analysis showed comparable expression of Ki-67 and p53 in tumors from the RAB27A-KO and RAB27B-KO mice compared to controls. However, MMP2 and MMP9 expression

Discussion
We previously reported on the frequent expression and clinical relevance of several molecules belonging to the exosome secretion machinery in HGSC cells and tumorderived exosomes [19]. In the present study, we analyzed whether a novel 3D OC model representing OC progression can be relevant for studying exosome secretion pathways in this cancer. We propose that alginate scaffolds emulate solid tumors, while spheroids mimic tumor growth in effusions. In the    present study we show that the 3D form of growth influences both the mRNA and protein expression of the exosome secretion pathway in OC cell lines, as reflected by changes in NSMASE2, TSAP6, RAB27A, and RAB27B mRNA, as well as NSMASE2 and TSAP6 protein expression depending on the 3D form applied.
In OVCAR8 cells, TSAP6 and RAB27A mRNA levels were higher in spheroids than in scaffolds and RAB27B mRNA expression was not influenced by the growth form.
is profile partly concurs with our findings in our previous study [19], where TSAP6 mRNA expression levels were higher in HGSC effusions compared to surgical specimens, while those of RAB27B expression did not differ between these anatomic locations. Data are discordant with respect to RAB27A mRNA, which had comparable expression levels in solid lesions and effusions, despite the use of the same assay, presumably due to inherent differences between cell lines and clinical specimens. NSMASE2 and TSAP6 protein levels were higher in scaffolds compared to spheroids, in agreement with the higher NSAMSE2 and TSAP6 expression in solid tumors versus effusions in our previous study. ese data suggest that the different culture models faithfully emulate 3D growth in solid lesions and effusions in HGSC.
Of note, there were notable differences between the mRNA and protein expression levels of Rab27A and TSAP6. We hypothesize this discrepancy to result from the action of molecules regulating mRNA levels or due to epigenetic regulation of protein expression.
We utilized the CRISPR-Cas9 system and produced NSMASE2, TSAP6, RAB27A, and RAB27B KOs in the OVCAR8 cell line. EM showed that OVCAR8 cells produce exosomes. e secretion of these exosomes was significantly diminished in RAB27A-KO, RAB27B-KO, and TSAP6-KO cells, well in agreement with the study of Bobrie et al., in which Rab27a mediated exosome secretion in breast carcinoma cells [22]. Of note, in another study by the same group, Rab27a silencing reduced docking of multivesicular endosomes (MVEs) and increased their size in Hela cells, whereas MVEs were redistributed towards the perinuclear region upon Rab27b silencing, documenting that these two proteins have different roles [23].
Functional assays in the present study showed that molecules involved in the exosome machinery also regulate critical cancer-related processes. In migration assays, all KO cells migrated significantly less, indicating that these genes, perhaps via exosome secretion, are needed for this metastatic feature. In the zymography assay that assesses the ability of    that is not amenable to surgical removal, where tumor cells are chemoresistant and possess cancer stem cell characteristics [24,25].
is form of tumor progression is therefore a highly lethal stage of OC, in addition to its detrimental effect on the quality of life of OC patients. In contrast to OVCAR8 controls, no development of ascites was observed following injection of RAB27A and RAB27B KO cells to mice, and solid tumor nodules were significantly smaller, indicating that these molecules may have direct clinical relevance in HGSC. Tumors from mice injected with RAB27A and RAB27B KO cells additionally had lower MMP expression. Overall, the experimental data, combined with the association between TSAP and poor outcome in our previous study [19], support a role as oncogenes rather than tumor suppressors for the exosome machinery components. An association between Rab27a protein and favorable outcome was nevertheless observed for exosomal levels of this protein in our previous study [19], raising the possibility that extracellular Rab27a has other functions than its intracellular counterpart.
In conclusion, this study presents a 3D in vitro model that may allow exploration of the molecular drivers of the different growth forms of OC. Our in vivo model shows that interfering with the exosome secretion pathway contributes to attenuating the aggressiveness of OC, a finding that may have therapeutic relevance. Analysis of the content of exosomes derived from the scaffold and spheroid model, including proteins, mRNA, and noncoding RNA, may be of interest in future research.

Data Availability
e experimental data used to support the findings of this study are included within the article. Additional information regarding this study is available from the corresponding author upon request.

Conflicts of Interest
e authors have no conflicts of interest.

Authors' Contributions
Esther Channah Broner performed the experiments and wrote the manuscript. Hadil Onallah performed the experiments and critically read the manuscript. Tali Tavor Re'em prepared the 3D constructs and critically read the manuscript. Ben Davidson designed the study, microscopically assessed the tumors in the mice experiments, performed and scored the immunostains, and co-wrote the manuscript. Reuven Reich designed the study, supervised the experiments, and cowrote the manuscript. Esther Channah Broner and Hadil Onallah contributed equally to this work.