Establishment and characterization of a cell line and patient-derived xenograft (PDX) from peritoneal metastasis of low-grade serous ovarian carcinoma.

Peritoneal spread indicates poor prognosis in patients with serous ovarian carcinoma (SOC) and is generally treated by surgical cytoreduction and chemotherapy. Novel treatment options are urgently needed to improve patient outcome. Clinically relevant cell lines and patient-derived xenograft (PDX) models are of critical importance to therapeutic regimen evaluation. Here, a PDX model was established, by orthotopic engraftment after subperitoneal tumor slurry injection of low-grade SOC, resulting in an early-stage transplantable peritoneal metastasis (PM)-PDX model. Histology confirmed the micropapillary and cribriform growth pattern with intraluminal tumor budding and positivity for PAX8 and WT1. PM-PDX dissociated cells show an epithelial morphotype with a 42 h doubling time and 40% colony forming efficiency, they are low sensitive to platinum derivatives and highly sensitive to paclitaxel (IC50: 6.3 ± 2.2 nM, mean ± SEM). The patient primary tumor, PM, PM-PDX and derived cell line all show a KRAS c.35 G > T (p.(Gly12Val)) mutation and show sensitivity to the MEK inhibitor trametinib in vitro (IC50: 7.2 ± 0.5 nM, mean ± SEM) and in the PM mouse model. These preclinical models closely reflecting patient tumors are useful to further elucidate LGSOC disease progression, therapy response and resistance mechanisms.

interest increases in the development of new treatment approaches 7 . Therapeutic management of ovarian cancer has traditionally been based on a combination of surgery and platinum-/taxane-based chemotherapy 6 . However, LGSOC is not as responsive to platinum-/taxane-based chemotherapy as HGSOC. The involvement of the MAPK pathway in the disease is also demonstrated in phase 3 studies by Grisham et al. and Gershenson et al. (2019). Both studies showed response rates to single agent MEK inhibitors, respectively binimetinib and trametinib, in patients with recurrent LGSOC 8,9 . In every aspect of translational cancer research, from the biological aspects of the disease to the development of new treatments, the use of preclinical models is a key component. In recent years, there has been an increasing interest in the application of organoids and patient-derived xenografts (PDXs) because of their high potential as an essential tool for personalized medicine [10][11][12] . The process of generating PDXs (also known as tumorgraft models) is based on the transfer of fresh tumor tissue (primary or metastatic) from the patient directly to an immunocompromised mouse 13 .
Depending on the cancer type, pretreatment, amount of tissue available, molecular properties etc., the success rate of the PDX will vary 14 . The organ environment can also affect tumor engraftment, highlighting the role of the site of implantation. Traditionally the tumor fragment is implanted into an area unrelated to the original tumor site, which is considered a heterotopic implantation (generally subcutaneous). On the other hand, tumor xenografts can also grow orthotopically into the corresponding anatomic region but their use is often hindered by a need for a high level of technical skills, time and cost 15 . For some cancers, such as colorectal, breast, lung, pancreatic, head and neck, melanoma, gastric, ovarian, prostate and renal cancer, methodologies for PDX establishment and characterization are already described in literature with engraftment rates ranging from 9 to 90% of success 14,16 .
In this work, for the first time, an orthotopic PDX model, based on a subperitoneal tumor slurry injection, and cancer cell line from a peritoneal metastasis of LGSOC were established. This model showed a KRAS mutation and sensitivity to the MEK inhibitor trametinib demonstrating its clinical relevance to study treatment responsiveness and resistance mechanisms.

Results
A low-grade serous ovarian carcinoma (LGSOC) peritoneal metastasis (PM)-PDX model. Figure 1A illustrates the establishment of the LGSOC PM-PDX model. Based on the observed increase in high-density signal from the ultrasound imaging (Fig. 1B), it was decided to passage the tumor tissue to a new group of acceptor mice 46 days after injection. This second passage was monitored over 3 months but no change in high-density ultrasound signal was demonstrated. At day 146 post implantation, a macroscopically blister-like appearance of the tumor area was observed which was formalin-fixed and processed for (immuno)histology. H&E revealed the micropapillary and cribriform growth pattern typical for LGSOC surrounded by a large mass of stroma in the first PDX passage. In the second PDX passage this micropapillary pattern dominated the tumor area showing a single layer of epithelium forming a large lumen (Fig. 1C). This micropapillary pattern was further characterized by intraluminal tumor budding. Immunohistochemical stainings for paired box gene 8 (PAX8), Wilms' tumor gene 1 (WT1), tumor suppressor protein p53, estrogen receptor (ER) and progesterone receptor (PR) further confirmed the typical characteristics of LGSOC (Fig. 1D). characterization of tumor-derived cell lines. Primary culture from single cell suspension of patient-derived peritoneal metastasis resulted in spread-polarized cells that typically showed signs of senescence characterized by a larger surface area and stress fibers ( Fig. 2A). These cultures showed a mixed expression of cytoskeletal proteins alfa-smooth muscle actin and cytokeratin and cell-cell adhesion molecules epithelial (E-) and neural (N-) cadherin, and most likely can be considered as mixed mesothelial-fibroblast cultures. In contrast, primary culture starting from tissue of the first passage PM-PDX model resulted in typical epithelial cells with cobblestone organization with strong cell-cell adhesion that showed colony growth. The first 5 to 8 initial subcultures showed no constant timing (among 2 to 3 weeks), the period in which cell proliferation was slow and unable to cover the entire culture flask surface. After this period, cell proliferation became quicker and in vitro passages for the maintenance of cell culture became regular (every week). The cell culture, named as PM-LGSOC-01, has been in continuous culture for >30 months and >100 in vitro passages ( Fig. 2A). PM-LGSOC-01 cells had a doubling time of 42 hours at passage 5 that was reduced to 23 hours at passage 22 and later passages. Table 1 summarizes the main findings regarding short tandem repeat (STR) analysis. Comparison of STR profiles between PM-LGSOC-01 and other human cell lines did not match evaluation values greater than 0.82, confirming the uniqueness of the PM-LGSOC-01 cell line. Results of western blotting ( Fig. 2B and Supplementary Figure S1) illustrate a stable expression of cytoskeletal and cell-cell adhesion proteins over a wide range of passage numbers. Despite the presence of E-cadherin and its associated cytoplasmic catenins the PM-LGSOC-01 cells form aggregates but do not show compact spheres within 48 h in contrast to positive controls used for compact sphere formation (Fig. 2C). Chemotactic migration of PM-LGSOC-01 to a 10% FBS gradient was limited in contrast to SK-OV-3 luc IP1 cells characteristically used as a migratory ovarian cancer cell line (Fig. 2D). PM-LGSOC-01 cells were considered being insensitive to estrogen signaling as no significant differences in colony formation were found between the different estrogen conditions and the control condition tested (Supplementary Figure S2).  Figure S3). Selumetinib also affected cell confluency but only in higher molar concentrations. In agreement with the poor chemosensitivity of LGSOC only paclitaxel shows a sensitivity in the low nM range (IC50 of 6.3 ± 2.2 nM (mean ± SEM)) in contrast to platinum based compounds with IC50 > 2 µM (Fig. 3C). In agreement, the clonogenic assay confirmed the effect of trametinib and selumetinib on clone numbers (Fig. 3C). Cell cycle analysis confirmed the impact of trametinib on cellular growth by stimulating a cell population into an increased G1 phase and decreased S and G2/M phase (Fig. 3D).

impact of trametinib in an in vivo peritoneal metastasis model of LGSOC. PM-LGSOC-01 cells
were lentiviral transduced to obtain constitutive EGFP and luciferase expression. These reporter cells were further used to create a peritoneal metastasis model from LGSOC in order to evaluate the effect of trametinib in vivo. Figure 4A illustrates the imaging data at different time points before and during the treatment period. In both groups, during the time course of the experiment no mice developed ascites. Animals received daily oral gavage based on vehicle or 0.3 mg/kg trametinib in a volume of 100 µL. Over time, a clear increase in bioluminescence activity can be observed for the control group whereas a decrease in signal is observed in the trametinib group. After 5 weeks of treatment animals were euthanized and relative total flux was significantly higher in the control group compared to the trametinib group (Fig. 4B). On average a 4-fold increase in bioluminescent increase from the start of the experiments was observed for the control group whereas on average the bioluminescent signal decreased with about 30% in the trametinib group, relative to starting conditions. Figure 4C illustrates the histopathological (H&E) and immunohistochemical stainings (Ki67 and PAX8) representative for both the control and trametinib group. H&E shows nests of cells that organize into papillae surrounded by stroma characteristic of LGSOC. Ki67 labeling index was twice as high in the control group (30%) compared to the trametinib group (15%).

Discussion
The heterogeneous nature of ovarian cancer makes it challenging to predict therapeutic responses in patients 17,18 . In this regard, preclinical models accurately mimicking biological properties of in vivo human tumors are of great value for efficient drug discovery 19 . To date, preclinical research in LGSOC has been limited. The low frequency and slow growth rate of these tumors have challenged the development of cell lines and animal xenograft models.
LGSOC cell lines are not available at the American Type Culture Collection (ATCC) and are only available at some research groups worldwide 20,21 . Kopper et al. established organoid lines in basement membrane extracts representing both LGSOC and HGSOC from primary tumor, ascites and peritoneal metastasis 12 . The organoid lines allow subcutaneous transplantation and can be used in drug screening assays. Our approach was slightly different. A tumor slurry from peritoneal metastatic tissue of LGSOC was subperitoneally injected into an immunodeficient SCID/Beige mouse leading to tumor growth. From this early-stage PDX a tissue-culture substrate adherent cell line was established that showed long term in vitro expansion and enabled manipulation and functional analysis. We also confirmed the histological features of the early-stage PDX such as micropapillae surrounded by stroma in the first passage and marked architectural complexity in the second passage most probably due to anastomosis of micropapillae forming the elongated and branching structures. The genomic aberration characterized by KRAS mutation is consistent in the PM-PDX and PM-LGSOC-01 cell line. Biomarker www.nature.com/scientificreports www.nature.com/scientificreports/ expression, such as positive PAX8 and WT1 combined with a wildtype p53 is consistent in the primary tissue versus the PM-PDX and PM-LGSOC-01, even after extended passage. Ovarian PDXs are predominantly originating from HGSOC as a low take rate and long latency is often associated with other histological subtypes. However, in our case HGSOC patients were strongly pretreated by chemotherapy and characterized by necrotic areas and areas containing cancer cells with low mitotic activity making it less likely to establish a PDX model from PM of HGSOC patients. In contrast, the LGSOC patient did not receive neoadjuvant chemotherapy before surgery leading to more viable tumor tissue, easily forming an early-stage transplantable PDX and generated tissue-culture substrate adherent PM-LGSOC-01 cell line, low sensitive to platinum derivatives. Other characteristics are clonogenicity and tumorigenicity, lack of serum-induced chemotactic migration and absence of compact sphere forming activity despite the presence of cell-cell adhesion molecule E-cadherin and its downstream catenins. The PM-LGSOC-01 cell line allows genetic manipulation and easy in vivo monitoring of its luc-EGFP variant using bioluminescence imaging. The mouse passaging of PM tumor tissue was necessary to obtain a tissue-culture substrate adherent cell line since cells cultured directly from patient PM tumor tissue ended into dominant growth of stromal cells such as fibroblasts and mesothelial cells that became senescent after further passaging.
Prior studies have reported that LGSOC tumors have a unique clinical, pathological and molecular profile compared to other ovarian cancers.
LGSOC harbours KRAS mutations in 19 to 54.5% of the cases and lacks TP53 mutations [22][23][24][25][26][27] . With the focus on inhibiting KRAS signaling via downstream effector MEK, both allosterically active compounds trametinib and selumetinib were here investigated 28 . Trametinib shows equal potency for targeting MEK1 and MEK2 and preferentially binds unphosphorylated MEK1/2 and thereby preventing RAF-dependent MEK phosphorylation and activation 29,30 . Selumetinib targets the unique inhibitor binding pocket adjacent to the Mg-ATP in MEK1/2. Attachment to this specific region causes a conformational change in unphosphorylated MEK1/2 resulting in a catalytically inactive position and blocking MEK1/2 from accessing the ERK1/2 activation loop. Selumetinib does not block binding and phosphorylation by RAF, which is different from trametinib 31 . In addition, selumetinib shows higher potency to target MEK1 compared to MEK2. These different binding properties of selumetinib compared to trametinib result in higher IC50 for selumetinib in MEK www.nature.com/scientificreports www.nature.com/scientificreports/ sensitive tumors (reported IC50 values of 50 nM for trametinib and 2.5 µM for selumetinib using the A549 bronchioloalveolar carcinoma cell monolayer cultures 32 ) which is in agreement with work done by Gilmartin et al. 29 and Yamaguchi et al. 33 . The study of Fernandez et al. 34 marks differences in MEK efficacy in LGSOC cell lines as trametinib was found to be highly effective in blocking p-ERK1/2 compared to selumetinib (IC50 values were in the nM range for trametinib versus the µM range for selumetinib). These findings are also in line with our observations regarding a different sensitivity for both MEK inhibitors with the established PM-LGSOC-01 cells. In vivo evaluation of trametinib in PM-LGSOC-01 revealed a similar sensitivity suggesting that the peritoneal stroma does not affect the trametinib response. Due to the ongoing evaluation of MEK inhibitors such as binimetinib 8 and trametinib 9 in phase 3 clinical trials for LGSOC, we strongly believe that this current model will assist in the better understanding of responsiveness and resistance to MEK inhibitors.
Establishing and analyzing additional LGSOC lines might substantiate our finding and may provide a unique opportunity to study LGSOC progression and chemosensitivity.

Methods establishment of peritoneal metastasis (pM)-pDX models.
Fresh peritoneal tissue specimens from 10 consenting patients with metastatic serous epithelial ovarian cancer (FIGO stage III or IV) were collected at the time of debulking surgery at Ghent University Hospital, Belgium. Nine patients were diagnosed with HGSOC and one with LGSOC. The study protocol was approved by the institutional review board of the Ghent University Hospital and the trial is registered as ClinicalTrial.gov NCT02567253 with EudraCT number 2015-000418-23. Samples were processed to a tumor slurry and injected in SCID/Beige mice within 75 minutes after removal from the patient. Tumors were minced in limiting volumes of RPMI 1640 media (Life Technologies, Ghent, Belgium), supplemented with 100 U/ml penicillin and 100 µg/ml streptomycin (Life Technologies, Ghent, Belgium). After a centrifugation step at 1500×g for 3 minutes, the upper culture medium was removed and tumor tissue was suspended in 1:1 Matrigel (Corning, The Netherlands). Further, a laparotomy was performed and 50 µL of the tumor suspension using a 19 G needle was injected bilateral subperitoneally in three female 4 to 5 week old SCID/Beige (C.B-17/IcrHsd-Prkdc scid Lyst bg-J ) mice (Envigo, The Netherlands). Animal studies were conducted in accordance with the local committee on the Ethics of Animal Experiments (Ghent University Hospital, Ghent, Belgium [ECD 15/28]). In order to cryopreserve tumors, tissue was minced and stored 1:1 in freezing media (90% FBS/10% DMSO) at −80 °C before storage in liquid nitrogen. tissue processing and immunohistochemistry. Tissues collected from mice or patients were fixed overnight in neutral buffered 10% formalin solution (Sigma-Aldrich, Belgium) and processed in the lab (H&E staining) or in the tissue core facility at Ghent University Hospital (immunohistochemistry). www.nature.com/scientificreports www.nature.com/scientificreports/ In vivo imaging. Transparent ultrasound transmission Polaris II gel (Ondes & Rayons Medical, France) was applied to bare skin and a MicroScan MS550D (22-55 MHz, VisualSonics Inc., Canada) transducer with the Vevo 2100 imaging system (VisualSonics Inc., Canada) was used to analyze the tumor cross-sectional area in Vevo LAB 1.7.1 (VisualSonics Inc., Canada).

establishment of tumor-derived cell lines.
To establish cell lines derived from the peritoneal metastasis and a PM-PDX model, tumor samples were cut into pieces of 2-4 mm and suspended using the tumor dissociation protocol with the GentleMacs dissociator (Miltenyi Biotec GmbH, Germany). The cell suspension was applied to a cell strainer (70 µm, Corning, The Netherlands), centrifuged at 300×g for 7 minutes and after aspiration of the supernatant resuspended in complete EMEM supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin and 100 µg/ml streptomycin (Life Technologies, Ghent, Belgium). The first weeks, cells were maintained in a 6-well plate (Novolab, Belgium) before culturing in a T25 falcon at 37 °C and 5% CO 2 in air. The cell culture was monthly tested for Mycoplasma by using MycoAlert Plus Kit (Lonza, Basel, Switzerland).
KRAS mutation analysis. Exons  incucyte ZooM monitored studies. Real-time monitoring of cell confluency was performed using the IncuCyte ZOOM System (Essen BioScience, Hertfordshire, UK) according to the manufacturer's guidelines. For cell confluency monitoring, cells were seeded in 96-well clear-bottom Corning Costar cell culture plates at 2 000 cells per well (100 µl/well) and allowed to adhere 24 hours at 37 °C and 5% CO 2 in air. Subsequently, cells were exposed to the drugs in complete EMEM medium and microscopic images (4 images/well) were taken every two hours for the duration of the experiment. All images were analyzed and cell confluency was deduced using IncuCyte software. Each condition was performed in, at least, four fold. Chemotaxis cell migration was studied for SK-OV-3 luc IP1 and PM-LGSOC-01 cells using the IncuCyte ClearView 96-Well Cell Migration plate coated with 1% Matrigel (Corning, The Netherlands) in 0% FBS EMEM medium. 3 000 cells/well were seeded (60 µL volume) with 0.1% FBS to the top and 200 µL 10% FBS to the bottom. Cell migration was followed using the phase contrast cell confluency monitoring. cell lysates and western blotting. Proteins were extracted from the cells using the Laemmli lysis buffer (0.125 M Tris-HCl, 10% glycerol, 2.3% sodium dodecyl sulfate (SDS), pH 6.8). After an ultrasonication step, cell lysates were suspended in reducing sample buffer (1 M Tris-HCl, 30% glycerol, 6% SDS, 3% β-mercaptoethanol, 0.005% bromophenol blue, pH 6.8) and boiled for 5 minutes at 95 °C. 20 µg proteins of the cell line were exposed to a 10% SDS-PAGE gel and transferred to nitrocellulose membranes (Bio-Rad, Hercules, CA, USA). After blocking the membranes using 5% non-fat milk or bovine serum albumin (BSA) in phosphate-buffered saline (PBS) with 0.5% Tween 20 (Sigma-Aldrich, Belgium), the membranes were incubated overnight at 4 °C with the primary antibodies (Table 2). After washing the membrane, incubation with HRP-conjugated secondary antibody was performed at room temperature for 1 hour. WesternBright Quantum HRP substrate (Advansta, Menlo Park, CA, USA) was added to the membranes to capture the luminescent signal using the Proxima 2850 Imager (IsoGen Life Sciences, De Meern, The Netherlands). Equal loading of samples was verified by primary monoclonal mouse anti-GAPDH antibodies (clone GAPDH-71.1, Sigma-Aldrich, Belgium).

Statistical analysis.
Results obtained with the colony formation assay were analyzed using one-way ANOVA with Tukey's multiple comparisons test using Graphpad Prism 7 (GraphPad Software, USA). Using R Studio 36 , Mann-Whitney U test was used to compare differences in bioluminescent signals (relative total flux) between groups in the in vivo experiment. Statistical tests were two-sided and p-value below 0.05 was considered statistically relevant. In figures, *represents p-value ≤ 0.05, **p-value ≤ 0.01 and ***p-value ≤ 0.001. ethical approval and informed consent. Informed consent of the patients to use tumor material was obtained after the study protocol was approved by the institutional review board of the Ghent University Hospital. Animal experiments were conducted in accordance with the local ethics committee (ECD 15/28, Ghent University Hospital).