Tumorigenicity of Ascites-Derived Tumor Cells: Insights Into the Molecular Mechanisms of Ovarian Cancer Progression and Therapy Resistance

Background: Ovarian cancer (OvCa) cells disseminate primarily intraperitoneally. Here, detached tumor cell aggregates (spheroids) from the primary tumor are generally regarded as “metastatic units”, which exhibit a survival benet, probably due to the protective microenvironment and their unique molecular characteristics. Hence, current therapeutic concepts such as classical chemotherapy are not sucient preventing growth and spread of OvCa spheroids. Methods: In the current study we analyzed the cellular composition of ascites from ovarian cancer patients using ow cytometry and the tumorigenic potential of the different subpopulations in an intraperitoneal mouse model. Comparative transcriptome analyses (RNAseq) from ascites-derived tumor cells spheroids (n=10) vs. tumor samples from different metastatic sites (n=30) were further performed in order to identify key molecular players responsible for the special cellular characteristics of OvCa spheroids. Results: In vitro culture of ascites-derived cells gave rise to two different subpopulations: an adherent cell population (ADs) including mainly CD90+ cells with highly proliferative rates in vitro but no tumorigenic potential in vivo, and a non-adherent cell population (NADs) containing principally EpCAM+/CD24+ cells with low proliferative potential in vitro. NADs included cell aggregates and single cells, the rst showing a high content (> 80%) of tumor cells (EpCAM+/CD24+). Enriched tumor cell spheroids from the ascites using cell strainers showed higher tumorigenic potential in vivo in comparison to the original ascites-derived cell population. Interestingly, the different metastatic spread patterns observed in the mice resembled the tumor dissemination pattern found in the corresponding patients. RNAseq analyses from tumor-spheroids revealed up-regulation of genes involved in chemoresistance (TGM1, HSPAs, MT1s), cell-adhesion and cell barrier (PKP3, CLDNs, PPL) and the oxidative phosphorylation (OXPHOS) process compared to the solid tumor tissue samples. Additionally, down-regulation of extracellular matrix components and angiogenesis-related

bloodstream, although pelvic and para-aortic lymph nodes are often involved. Frequently, patients develop ascites, a pathological uid within the abdominal cavity containing tumor cells but also cells from no tumorigenic origin and diverse soluble factors that create a favorable environment for tumor growth and dissemination. Tumor cells within the ascites are present either as single cells or more frequently as aggregates or so-called oating spheroids (2,3), the last representing the main source for peritoneal metastasis. Also, in advanced ovarian cancer patients without ascites, tumor cell spheroids are also present in the peritoneal lavage uid collected during surgery. Recently, using an in vivo ovarian cancer model, it could be shown that tumor spheroids mainly arise from multicellular detachment from the primary tumor rather than from single cells aggregating within the abdominal cavity. More precisely, cell aggregates in ascites arise from collective dissemination of neighboring cells in the primary tumor and in turn single spheroids developed individual metastatic sites (4). These observations are in line with the genetic clonal mapping of disseminated ovarian cancers in patients. Here, phylogenetic analysis of multiple intraperitoneal sites from individual patients showed that most metastatic sites were phylogenetically pure (5). Tumor cells within cell spheroids exhibit a survival advantage in comparison to single oating cells due to the protective microenvironment created by cellular interactions with other tumor and non-tumoral cells. Recent in vivo data suggest that tumor-associated macrophages (TAMs) may promote spheroid formation and tumor growth. Here, TAMs were found within the spheroids, thereby secreting large amounts of epidermal growth factor (EGF), which leads to upregulation of integrin and intercellular adhesion molecule 1 (ICAM-1) expression in tumor cells (6). Furthermore, spheroids represent a more chemoresistant population since chemotherapeutic drugs show a lower incorporation and poor diffusion in such multicellular structures (7,8). Moreover, it has been suggested that chemoresistance is caused, in part, by tumor cells entering a non-cycling state with low metabolic activity, which is a characteristic feature of detached tumor cells and spheroids oating in the peritoneal uid or ascites (3).
In conclusion, ovarian carcinomas are highly heterogeneous tumors and tumor cell aggregates (spheroids) that detach from the primary tumor comprise unique clones, which might or might not survive in the peritoneal uid. We assume that those spheroids that persist and are able subsequently to attach to the peritoneal cavity can be considered as "metastatic units". These cellular structures share certain biological characteristics that might be useful in the development of new therapeutic strategies.
In the present study we aim to characterize the cellular composition and the tumorigenic potential of the different subpopulations found in the ascites uid of ovarian cancer patients. Further, a rst insight into the speci c molecular characteristics of tumorigenic ascites-derived tumor cell spheroids was accomplished by RNAseq analyses.

Patient material
Ascites was collected from patients diagnosed with advance ovarian cancer at the University Medical Centre Hamburg-Eppendorf between 2017 and 2020. Ascites was obtained during debulking surgery from patients with primary and recurrent disease. Detailed patient characteristics are presented in supplementary table 1. All patients gave written informed consent to access their biomaterial and review their clinical records according to our investigational review board and ethics committee guidelines (#190504 and PV6012). Clinical parameters were retrieved from a detailed institutional database providing information on clinicopathological factors, surgical and therapeutic procedures.
Preparation, cultivation of ascites-derived cells and isolation of ascites-derived spheroids Ascites or lavage from advanced ovarian cancer patients were centrifuged at 1200 rpm for 5 minutes at room temperature. Supernatants were collected and frozen down at -20 °C for other purposes. Cell pellets were resuspended in RBC lysis buffer (Red Blood Cell lysis buffer, Invitrogen, San Diego, CA, USA) and incubated for 15 minutes at room temperature. After 5 minutes centrifugation at 1200 rpm, the cell pellets were washed with PBS (Phosphate Buffered Saline solution, Sigma-Aldrich, St. Louis, Missouri, USA) and resuspended in PBS. 10 µl of cell suspension were mixed with 10 µl of a 0.04% trypan blue solution and cell amount, cell size and aggregation status were observed under the microscope.
Ascites-derived cells were cultured in MCDB medium (MCDB 105 Medium and Medium 199 (1:1) supplemented with 10% fetal calf serum (FCS), and penicillin/streptomycin (2 mM) (1%) (Life Technologies, CA, USA) or in the clari ed supernatant of the ascites at 37 °C on low attachment plates in the presence of 5% CO 2 and 95% humidity. Here, some cells oated as spheroids or single cells in the medium (non-adherent cells, NADs) while some cells attached to low attachment plates (adherent cells, ADs).
For some experiments, large-sized cell aggregates present in the ascites-derived pellet were separated by using 15 µm cell strainers (pluriSelect, Leipzig, Germany). Brie y, the cell pellet was resuspended in PBS and put on the top of the cell strainer softly shanking until no uid in the upper part was observed.
Additional 5-10 ml PBS were added to the cell strainer to ush and wash the cells. The uid containing all cells < 15 µm were collected in a 50 ml tube. Subsequently the cell strainer was turned upside down and cells and cell aggregates retained in the strainer were ush back onto a new 50 ml tube using 5-10 ml PBS.
In vivo intraperitoneal mouse model Ascites-derived cells (aprox. 3 × 10 6 cells) from the original ascites cell pellet or from the separated tumor spheroid fraction were resuspended in MCDB medium without FCS (200 µl) and injected into the peritoneal cavity of immunode cient female mice (CB17/Icr-Prkdcscid/IcrIcoCrl (SCID) or C.129S6(B6)-Rag2tm1Fwa N12 (Rag2-Model 601), as previously described (9). Due to the limited number of ascitesderived cells available after preparation, habitually one mouse was injected per patient sample. For some patient samples, injections with cells corresponding to different size fractions were performed. In total 30 mice were included in this study. The animals were housed with a 12 h day-night cycle in a temperature-(21 °C) and humidity-(50%) controlled room. All mice were kept in individually ventilated cages under pathogen-free conditions, fed with sterile standard food and water ad libitum. Mice that showed strong signs of tumor progression (ascites, shaggy coat and loss of appetite; (10)) were anesthetized with xylazine/ketamine (120/16 mg/kg body weight, Bayer, Leverkusen, Germany/ Graeub, Bern, Switzerland) and sacri ced after terminal cardiac blood collection by cervical dislocation. The dissemination pattern found was documented, tumors at the injection site, metastases and lungs were excised, frozen or formalin-xed and embedded in para n. Animal experiments were conducted according to the UKCCCR guidelines for the welfare of animals in experimental neoplasia (10). The mouse experiments were approved by the local licensing authority (Freie und Hansestadt Hamburg, Behörde für Gesundheit und Verbraucherschutz, Amt für Verbraucherschutz, project #G16/55).

Results
Characterization of cellular components in ascites from ovarian cancer patients The ascites or lavage from OvCa patients (n = 141) were collected during debulking surgery. After the rst centrifugation step, only samples containing a visible cell pellet were morphologically assessed by phase contrast microscopy immediately after collection on day0 (n = 75). In 60% of the ascites samples we found both single cells and cell aggregates (spheroids) as shown exemplary in Fig. 1  Further, cell strainers of 15 µm were used to separate the original pellet into a small and a big cellular fraction. The multicellular aggregates (spheroids) were mainly collected in the big fraction, while the cell population smaller than 15 µm, de ned as small fraction, included only single cells (Fig. 1). After this ltering step the CD24+/EpCAM + cell population was strongly enriched in the big fraction compared to the original and small pellet, while most CD45 + and CD90 + cells were found in the small fraction, which contained a relatively low amount of CD24 + and/or EpCAM + single cells.

Cellular characterization of different subpopulations arising from ascites-derived cells in vitro
Ascites-derived cells (day0) were cultured in low-attachment plates with MCDB medium for up to two weeks. Here, two different cell populations could be distinguished after approx. 5 days cultivation time: adherent elongated spindle-shaped mesenchymal-like or cobblestone-shaped epithelial-like cells (adherent cells, ADs) and multicellular aggregates (spheroids) and small single cells oating in the supernatant without attachment (non-adherent cells, NADs) (Fig. 3A). Both cell populations showed a remarkably different proliferative behavior. ADs were typically highly proliferative, whereas the NAD population remained quiescent but viable for up to two weeks cultivation time. In order to better characterize these two cell populations, an additional FACs panel including a tumor marker (EpCAM), stem cell markers (CD44, CD24 and CD133) and mesenchymal-like and mesothelial cell markers (CD90, podoplanin and mesothelin) was established. Thus, the AD population (n = 9) exhibited a high content of CD90+, podoplanin + and to some extent mesothelin + and CD44 + cells, whereas NADs (n = 9) were mainly EpCAM + and CD24+ ( Fig. 3B and Supplementary Fig. 1). Figure  These ndings led us to the conclusion that in most of the ascites-derived samples the highly proliferative population of adherent cells (ADs) does not comprise ovarian cancer cells but rather a mesenchymal-like cell population and cells from mesothelial origin. To prove this assumption, the tumorigenic potential of these cells were further tested in vivo using an intraperitoneal mouse model.

In Vivo Tumorigenicity Of Ascites-derived Cells
In a next step we were interested in the tumorigenicity of the different populations found in the ascitesderived cells in vivo. Therefore, ADs, spheroids (cell fraction size > 15 um), single cells (cell fraction size < 15 um) and/or original pellets from different ovarian cancer patients (n = 21) were each intraperitoneally injected in immunode cient mice. None of the mice injected with ADs (n = 7) developed peritoneal carcinomatosis, whereas injection of the original ascites cell-pellets, without cell separation or previous cultivation, led to tumor or ascites formation in 50% and 20% of the injected samples, respectively. Similarly, all spheroid samples (n = 9) gained from the original pellet via ltration with a cell strainer developed in a time period ranging from 18 to 56 weeks metastatic ovarian cancer, thereby showing different extents of dissemination and patterns (Fig. 4). For two cases, the corresponding single-cell fraction was injected in parallel. Here, no tumor formation could be observed, even though this fraction contains single tumor cells, as showed before. Interestingly, we found a strong similarity in the tumor dissemination pattern observed in mice and in the corresponding ovarian cancer patient, as described in the surgery protocol. In our mouse model, we could clearly distinguish between a miliary-like dissemination pattern found in 11 samples and a "non-miliary" tumor spread generated by 2 samples, the last characterized by one or two big tumor bulks within the peritoneal cavity or retroperitoneal located (n = 2). One sample developed only malignant ascites, containing big amounts of tumor cell spheroids without any solid tumor lesion. Remarkably, reinjection of mouse ascites-derived tumor cells led to tumor and to some extent to ascites development in all cases (n = 2), thereby maintaining the same spread pattern and showing faster progression rates.

Transcriptome Of Tumor Cell Spheroids From Ovarian Cancer Patients
As shown in the in vivo model, ascites-derived tumor cell spheroids from ovarian cancer patients can be considered "metastatic units" that promote peritoneal carcinomatosis as well as retroperitoneal tumor cell dissemination. We assume that tumor cell spheroids biologically differ from those tumor cells in the primary or metastatic tumor tissue, since they must be able to survive as oating units and eventually to attach at the metastatic site. In order to identify the molecular players responsible for these speci c biological characteristics, the transcriptome of ascites-derived tumor spheroids and tumor tissue samples were compared. RNAseq analysis from 10 different tumor spheroid patient samples and 30 tumor tissue samples, including primary and metastatic tissue, were performed. Here, the ascites-spheroid (A1-A10) and the tumor sample group (T1-T30) form two distinct clusters in the principal component analysis (PCA; Fig. 5A). Subsequently, comparative expression analyses between the ascites group and three different tumor groups, each one containing 10 tumor samples, were carried out (Fig. 5B). The tumor tissue samples included in each of the three groups were selected based on the patient characteristics (FIGO, age, histology, grading, tumor stage and lymph node status) to resemble the ascites group. Only those genes commonly and signi cantly (|Log 2 -fold change| > 1 and FDR < 0.1) de-regulated in the ascites-derived spheroids in comparison with all three tumor groups were further evaluated and are displayed in the supplementary tables 2 and 3. A selected group of these genes has been presented in the heat map on Fig. 5. Interestingly, the oxidative phosphorylation pathway, including a large number of genes, i.e. ATPsynthases, NADH oxidoreductases and cytochrome c oxidases, are signi cantly upregulated (p value: 5.6952e-12, FDR: 1.6539e-8) in the ascites spheroids in comparison to the tumor cells from the primary or metastatic tissue. Higher mRNA levels of genes related to chemoresistance (i.e TGM1, heat shock proteins and metallothioneins), to cell-cell adhesion and also barrier molecules (i.e. PKP3, PPL, CLDN4/7 and laggrin) were found in ascites spheroids vs. tumor tissue as well. Three glycosylation enzymes, NEU1, B4GALT5 and CHST4 as well as several transcription factors (FOS, JUN and KLF4/6) were also up-regulated in the ascites group. Interestingly, the increased expression of CD163 and MARCO (macrophage receptor with collagenous structure) in the group of ascites-derived spheroids, suggest an important role of macrophages on the biology of these cellular structures. Among the 1316 signi cantly down-regulated genes in the ascites spheroids, two main pathways: angiogenesis and extracellular structure organization could be identi ed, which are signi cantly down-regulated (p value and FDR are almost 0). Numerous genes involved in angiogenesis (i.e ANGTP2, APLNR, PDGFRß or the cytokines CCL11 and CCL2) and as expected in the extracellular structure organization (i.e. several collagen proteins, MMP16/19, bronectin, lumican, versican, and the hyaluronan synthase 2).
Inhibition of the OXPHOS pathway via metformin treatment in tumor spheroids from ovarian cancer patients In order to evaluate the effect of an OXPHOS pathway inhibition on ascites-derived tumor spheroids from ovarian cancer patients, in vitro cell viability was measured after treatment with metformin alone as well as in combination with cisplatin. Here, 5 samples from different ovarian cancer patients were incubated with increasing metformin (1 and 5 mM) and cisplatin concentrations (1, 5 and 10 µg/ml) for 48 hours. Cell viability was assessed subsequently using CellTiter-Glo solution as described in the methods section. Among the 5 samples analyzed, 3 samples showed a strong response to the metformin treatment, even at the lowest concentration of 1 mM. In these cases the simultaneous treatment with metformin and cisplatin led to a signi cant viability reduction when compared with each treatment alone. One sample (#11) showed no response to metformin, whereas it was highly sensitive to cisplatin treatment. Interestingly, one sample (#15) from a recurrent ovarian cancer patient showed as expected no response to cisplatin treatment, but a signi cant reduced cell viability after 48 treatment with 1 and 5 mM metformin. The simultaneous treatment with cisplatin and metformin led surprisingly to an even stronger cytotoxic effect on this ovarian cancer sample (Fig. 6).

Discussion
In ovarian cancer, detached single and tumor cell aggregates (spheroids) from the primary tumor that persist in the peritoneal uid represent the main source of intraperitoneal metastasis (14). Specially, tumor cells within the spheroids exhibit a survival bene t and may represent a key element of chemotherapy-sensitive recurrence. In the present study we were able to identify and enrich this tumorigenic subpopulation within the ascites of ovarian cancer patients and further elucidate via RNAseq analysis unique molecular characteristics of these cellular structures.
Malignant ascites itself constitutes a favorable milieu for tumor cells to progress. It contains soluble factors such as cytokines, chemokines, growth factors, and extracellular matrix fragments as well as a complex mixture of cells including tumor, stromal cells and in ltrating immune cells [2]. The cellular part of ascites includes single cells and cell aggregates, so-called oating spheroids [5,6]. In the present study, we could show that the single cell population includes some tumor cells, but it is principally composed of immune cells and to a less extent of a mesenchymal-like cell population, de ned in our analysis by a CD90 positivity. In contrast, the cell aggregates found in most ovarian cancer samples contained a much higher (> 80%) percentage of tumor cells showing a strong EpCAM expression or combined EpCAM and CD24 positivity. Here, the tumor purity as well as the protective environment created by these structures may explain the high rate of successful tumor development observed, when injecting tumor-spheroids intraperitoneally in immunode cient mice, in contrast with single cell tumor injections. Moreover, the interaction between tumor cells and other cellular components within the spheroids seems to be essential in order to keep their compact structure, but it also enhances the survival ability and invasive potential of the tumor cells. In this context, different cell populations such as cancer-associated broblasts (CAFs) or tumor-associated macrophages (TAMs) have been described as key players for the aggregation as well as for the adhesive and invasive properties of these tumor cell structures, thereby potentiating their malignant phenotype and facilitating the peritoneal metastatic process (15)(16)(17)(18). Our FACs and RNAseq analysis on puri ed tumor spheroids are in line with this ndings. We could frequently detect CD45positive and CD90-positive cells within the spheroids, although both represented a small fraction within the spheroids, and further two macrophage-associated genes (CD163 and MARCO) were found to be highly up-regulated in the tumor-spheroids in comparison with the solid tumor tissue. CD163 is a characteristic marker of M2 macrophages, which are the most predominantly TAM subtype found in ovarian cancer, and are associated with tumor invasion, angiogenesis, metastatic disease and early recurrence (18)(19)(20). MARCO is a class A scavenger receptor expressed by immune suppressive tumorassociated macrophages and has been linked to poor prognosis in breast cancer (21,22). Interestingly, targeting MARCO-positive TAMs with a speci c antibody reduces tumor growth and metastasis in breast, colon and melanoma mouse models (23). In glioblastoma, MARCO-expressing TAMs induce a phenotypic shift towards mesenchymal cellular state of glioma stem cells, promoting both invasive and proliferative activities, as well as therapeutic resistance to irradiation (24). Additional analysis are required in order to elucidate the impact of MARCO-positive TAMs on spheroid tumor cells and whether this interaction might further in uence disease progression of ovarian cancer patients.
One characteristic feature of tumor spheroids is their low chemosensitivity, in part attributed to a low proliferative pro le (3). The standard chemotherapy for ovarian cancer patients, consisting in a paclitaxel and carboplatin combination, selectively targets and eliminates highly proliferative tumor cells (25). In poorly vascularized tumor areas however, cells become quiescent (26) and in turn less responsive to therapy. A recent study has estimated that in ovarian cancer spheroids more than 60% of the cells are quiescent (4). Moreover, it has been described that quiescent tumor cells use preferentially the mitochondrial OXPHOS pathway for their ATP production (27). In line with this data, our RNAseq analysis revealed a signi cant up-regulation of the OXPHOS pathway in the tumor-spheroids isolated from the ascites of ovarian cancer patients compared to corresponding solid tumor tissue samples. Thus, OXPHOS pathway inhibition opens an attractive therapeutic window for the speci c target of tumor spheroids, as the major vehicle of peritoneal metastasis in OvCa. By using OXPHOS inhibitors, spheroid tumor cells might not be able to cover their high ATP demand. Contrary to normal cells that can activate glycolysis in response to OXPHOS inhibition, quiescent tumor cells within the spheroids have no access to su cient glucose in order to compensate the loss of ATP production and might die (28). Interestingly, several drugs, including metformin, that have been used clinically for non-oncologic indications have emerged as effective OXPHOS inhibitors (29). Several cohort studies have described a protective effect and an association of metformin with longer overall survival in ovarian cancer patients (30)(31)(32)(33). Also, two in vitro studies have found reduced ovarian cancer cell proliferation, migration and increased apoptosis (34) as well improved sensitivity in drug-resistant ovarian cancer cell lines (35) after metformin treatment. In contrast, a recent meta-analysis that excluded studies considered to have the potential for immortal time bias, suggested no overall survival bene t associated with use of metformin (36). A recent pilot study evaluating the e cacy of metformin plus rst-line chemotherapy versus chemotherapy alone in a small cohort of ovarian cancer patients found no effects of metformin, neither(37) Our in vitro analyses might explain these contradictory ndings. Here, ascites-derived tumor-spheroids from different ovarian cancer patients showed in the majority of the samples a strong effect of metformin alone as well as in combination with cisplatin treatment, on the tumor cell viability, including a cisplatin-resistant tumor from a recurrent ovarian cancer patient. One sample, however, showed no response to metformin even at high concentrations, indicating that inhibition of the OXPHOS pathway might be not a universal target for all ovarian cancer patients. Ongoing analyses in our group aim the identi cation of the speci c molecular features that might discriminate metformin responders from non-responders.
Our RNAseq data has further revealed several factors up-regulated in spheroids that are linked to cell chemoresistance. The transglutaminase 1 (TGM1), an enzyme that is mainly found in the epidermis, catalyzes protein bonds, so-called cross-linking, which give the tissue strength and stability. In gastric carcinoma TGM1 has been shown to promote the stem cell character and chemoresistance of tumor cells via modulation of the Wnt/beta-catenin signalling pathway. Further, several members of the heat shock protein 70 family (Hsp70) were found to be signi cantly up-regulated in the tumor spheroids compared with the tumor tissue. The human Hsp70 family consists of eight highly homologous members of chaperone molecules that differ in their intracellular localization and expression pattern. Specially, HSPA1A/1B and HSPA6, which code for the proteins Hsp70 and Hsp70-6, respectively are only expressed at low or undetectable levels under physiological conditions, but are rapidly induced by cellular stress (38). In cancer cells, the effect of Hsp70 has been not only related to its chaperone activity, but rather to its antiapoptotic role and the regulation of cell signaling. In ovarian cancer, increased Hsp70 expression was found in chemoresistant cells. Here, Hsp70 proteins block the translocation of Bax into the mitochondria and the release of mitochondrial proteins into the cytosol (39). Additionally, three metallothioneins (MT1E, MT1M and MT1X) were highly up-regulated in the spheroids. MTs are small cysteine-rich proteins with a key role in metal homeostasis and protection against heavy metal toxicity. Consequently, a drug resistance function has been described in the context of cancer (40), though speci cally in ovarian cancer no difference between MT expression in tumors from chemotherapy-treated vs. untreated patients could be found (41). Still, MT expression has been negatively associated with survival time in primary ovarian carcinomas (42).
Besides the low proliferative rate and chemosensitivity, tumor cells within the spheroids might acquire speci c adhesive characteristics that support a protective and compact cellular aggregation structure (43). In this context, our RNAseq analysis revealed high mRNA levels of integrin α3 (ITGA3), claudins 4 and 7 (CLDN4 / 7), desmosome proteins plakophilin 3 (PKP3) and periplakin (PPL) as well as the barrier protein laggrin (FLG), the last showing an aprox. 10fold up-regulation in the ascites-derived spheroids compared with tumor tissue. Interestingly, recent data raised the possibility that molecules with mechanical barrier function may be used by cancer cells to protect them from immune cell in ltration and immune-mediated destruction. Here, authors identi ed eight genes, including PPL and PKP3, whose increase expression in human melanoma metastases and ovarian cancers was associated with a lack of Th1 immune signatures and further strongly correlated with shorter overall survival (44).
The in vitro and in vivo behavior of ascites-derived cells has been reported by other groups before (2,43,45,46). In our study we showed similar results as previously described, namely in the majority of the samples ascites tumor spheroids from ovarian cancer patients showed a quiescent and non-adhesive phenotype when cultured in vitro, whereas the single cells gave rise to an adherent and highly proliferative population. In contrast, intraperitoneal injection in immunode cient mice showed just the opposite picture; namely tumor development was observed in most of the spheroid samples, but none of the ADs developed carcinosis. FACS and ICC analyses revealed an explanation for this contrary behavior, showing that the spheroids consist principally of tumor cells population, whereas the single cells were mainly of non-epithelial origin. We assume the lack of an adequate stimulus in vitro prevent tumor spheroids to attach and further spread, thereby highlighting the key role of the intraperitoneal environment for tumor progression in ovarian cancer. Thus, the key role of broblasts, immune, adipocytes, mesothelial and endothelial cells for disease progression has been broadly described in the last years (47, 48).
Remarkably, two samples showed a totally different pattern regarding their cellular distribution and in vitro behavior. Here, the AD populations included a high percentage of tumor cells, as they showed a strong staining for EpCAM. These results emphasizes the high heterogeneity of "ovarian cancer" and the need to decipher the different biological subtypes behind this entity, in order to develop speci c and targeted therapies. In this context, the ascites-derived tumor spheroids might represent a suitable model to address this question. In the present study we could show that ascites-derived spheroids from ovarian cancer patients clearly depict the biology of the individual disease, especially if we consider that their dissemination pattern in the mice clearly mimic the one observed in the patient.
Moreover, we could identify several molecular players that might help us to better understand the biology of ascites-derived tumor spheroids. Ongoing analysis in our group aim to prove the functionality of these markers and to explore their role as potential therapeutic targets.