Proteomic analysis of malignant ascites and its impact on ovarian cancer spheroids Authors: Jack Scanlan, Parul Mittal, Jennifer Washington, Clifford Young, Michael Antoniou, Marina Kochetkova, Martin K. Oehler, Peter Hoffmann, Manuela Klingler-Hoffmann Published: 2026 ABSTRACT BackgroundMost advanced ovarian cancer patients develop malignant ascites, which describes a buildup of fluid in the peritoneal cavity caused by increased vascular permeability and obstructed lymphatic drainage. Malignant ascites contains cancer cells, which can aggregate as spheroids, as well as stromal cells, cancer-associated fibroblasts, and blood cells that create a complex tumor microenvironment. This study explores the proteome of ascites and how this environment affects the viability, phenotypes, proteomes, and treatment responses of ovarian cancer cells.MethodsUsing label-free proteomics, we compared the proteomes of cell-free malignant ascites from ovarian cancer patients with those of serum. Additionally, we examined the ex vivo chemotherapy responses of cancer spheroids cultured in ascites. Through detailed proteomic analysis of cells grown as 2D or 3D in tissue culture medium or ascites, we identified biological pathways and specific proteins induced by ascites. Finally, we performed orthogonal validation of a candidate marker, TGM2, using immunofluorescent staining.ResultsProteomics of cell-free ascites identified increased levels of extracellular, secreted, and membrane proteins when compared to serum. Ascites enhanced the baseline cell viability and spheroid formation of immortalized ovarian cancer cell lines compared to standard cell culture medium. However, chemotherapy-induced cell death of spheroids grown in standard cell culture medium remained proportional to changes observed in ascites-cultured spheroids. Ascites-driven phenotypic changes were not recapitulated by adding selected chemokines nor periostin to the cell culture medium, suggesting that additional factors are required. Notably, ascites induced similar ECM, secreted, and membrane proteins across 2D and 3D models, including TGM2, which was validated in spheroids through immunofluorescent staining.ConclusionsThis study contributes to our understanding of the role of ascites in the regulation of the proteome and viability of cancer cells. It provides evidence for the induction of TGM2 expression by ascites. Results from this pilot study warrant further study in a larger cohort.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12014-026-09619-y. Background Ovarian cancer remains the most lethal gynecological malignancy [1], primarily due to advanced-stage diagnosis and innate or acquired chemotherapy resistance. Despite advances in surgical techniques and chemotherapy regimens, the five-year survival rate of advanced ovarian cancer patients has remained below 50 per cent for several decades [2]. Ovarian cancer often correlates with the presence of malignant ascites, which describes the accumulation of fluid within the peritoneal cavity [3]. While most patients with advanced disease accumulate more than 2 L of ascites [4], even a small volume of ≥ 100 mL can indicate poor prognosis [5]. Malignant ascites contributes to patient discomfort and morbidity and creates a complex tumor microenvironment due to the presence of aggregated cells, known as spheroids, that shed from the primary tumor [6]. Malignant ascites can also contain immune cells, extracellular vesicles [7–9], cytokines, growth factors, enzymes, and cell-free DNA [10], which have been shown to contribute to cell survival, proliferation, and dissemination of cancer cells [11, 12] enabled through direct contact with the peritoneal lining [13]. Proteomic and metabolomic profiling techniques have expanded our view of the ascitic milieu and have highlighted its role in disease progression. For instance, altered T-cell function, interleukin pathway upregulation, and reduced MHC-II expression reduce anti-tumor and immunosuppressive effects to promote disease progression [14–16]. Additional factors within ascites can also promote metastasis; for instance, through TGF-β1-mediated increases to the proliferation and migration of mesothelial cells [17]. Comparative studies of cirrhotic and malignant ascites have also identified cancer-specific pathways, metabolites, and spliceosomal components [18] in the ascitic tumor microenvironment. The identification of altered biological pathways and diagnostic and prognostic markers of ovarian cancer support the use of mass spectrometry-based proteomics in personalized medicine approaches [19]. For instance, a 64-protein signature can reliably forecast whether high-grade serous ovarian tumors respond to standard therapies with 98 percent specificity [20]. While more limited, some studies have recognized the value of malignant ascites as a window into the evolving molecular landscape of individual ovarian cancers, enhancing the detection of candidate biomarkers that could be pivotal in early diagnosis and the personalization of treatments. For example, an early gel-based MALDI-TOF/TOF mass spectrometry study identified ceruloplasmin as a marker of treatment resistance in cell-free ascites and validated this finding by ELISA in 28 patients [21]. Using advanced data-independent acquisition mass spectrometry, our laboratory recently demonstrated that proteome profiling of patient-derived ascitic spheroids can distinguish clinical treatment responses [22]. However, the poor alignment of analytical techniques, sample types, treatment regimens, and reporting standards between such studies [19] have hindered the generalizability of findings. The conserved genetic landscape of ascitic material compared to primary ovarian tumors [23], as well as their influence on chemoresistance and metastasis, make patient-derived ascites an attractive tool with which to personalize treatment. Retrospective ex vivo treatment response assessments on three-dimensional ascitic cell cultures have recapitulated patients’ clinical responses to platinum-based chemotherapy [24–29], paclitaxel and inhibitors of PARP [25], and monoclonal antibody treatments [28]. Furthermore, prospective studies have used such data to successfully guide treatment selection [30, 31]. However, the myriad pro-tumorigenic factors within malignant ascites call into question the applicability of standard cell culture medium for use in predictive assessments of chemotherapy response. This is pertinent, given that alterations in growth conditions of ascites cells in adherent versus spheroidal cultures have enabled the identification of several differentially regulated proteins, such as the cell cycle arrest marker Aurora kinase B [32]. Despite being a valuable biofluid for the analysis of ovarian cancer, our understanding of ascites composition and utility in ex vivo chemotherapy response testing remains limited, as does its impact on spheroid formation. Here, we use morphological and mass spectrometry-based proteomics methods to investigate how the ascitic microenvironment affects spheroid aggregation and protein expression. Comparing this complex biofluid to healthy patient serum reveals notable differences in the levels of ECM, secreted, and membrane proteins. We also observe variations when comparing ascitic fluid from patients with different capabilities to promote spheroid formation ex vivo, which cause distinct morphological changes to immortalized ovarian cancer cell lines. Our findings show that barrier-forming cornification and keratinization pathways are significantly enriched in non-spheroid-promoting ascites, whereas collagens and laminins are more prevalent in spheroid-promoting ascites. Using both adherent and spheroidal cultures, we demonstrate that ascites induces the upregulation of the spheroid-promoting TGM2 protein, which is validated with immunofluorescent (IF) staining. Methods Cell culture The COV-362 (RRID: CVCL_2420) serous ovarian cancer cell line and SKOV-3 (RRID: CVCL_0532) adenocarcinoma cell line were purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA) and authenticated via a short tandem repeat (STR) DNA profile in October 2022. Mycoplasma-free and authenticated A2780 (RRID: CVCL_0134) ovarian endometroid and TYK-nu (RRID: CVCL_1776) adenocarcinoma cell lines were purchased from Cell Bank Australia, Westmead, NSW, Australia. SKOV-3 and COV-362 cells were grown in DMEM medium and A2780 and TYK-nu cells were grown in EMEM medium (Sigma Aldrich, MO, USA). All were cultured with 10% fetal bovine serum (Bovogen Biologicals, East Keilor, Vic, Australia) supplemented with 1% penicillin/streptomycin and 1% L-glutamine (Sigma Aldrich, MO, USA). Ascites collection Malignant ascites samples were collected from patients treated for ovarian carcinoma at the Royal Adelaide Hospital (RAH; Adelaide, Australia), with ethics approval from the RAH (#140,201) and Central Adelaide Local Health Network Human Ethics Committees (#R20181215) with patient consent and approval. Primary HGSC culture Primary cells were isolated from the ascites of advanced stage HGSC patients (n = 3). All primary cells were grown in Advanced RPMI 1640 medium (Life Technologies, Mulgrave, Vic, Australia) supplemented with 4 mM L-glutamine, 10% FBS (Sigma Aldrich, MO, USA), and antibiotics (100 U penicillin G, 100 µg/mL streptomycin sulphate, and 100 µg/mL amphotericin B; Sigma Aldrich, MO, USA). Clinicopathological details are found in Supplementary Table 1. Cells were maintained at 60–80% confluency for 3 passages prior to plating into 10 cm culture dishes. To estimate total cell numbers at harvest, a parallel dish was seeded with the same number of cells and counted manually. Following aspiration of the media, adherent cells were washed three times with 3 mL of pre-warmed PBS. Cell pellets containing up to 1 × 10⁷ cells were collected and washed three times with cold PBS, then stored at − 80 °C prior to lysis. Spheroidal cell aggregates were generated from primary cells in plasma-treated flat-bottom plates and poly-HEMA–coated opaque ultra-low attachment plates. Primary cells were cultured in 0%, 1%, 5%, 10%, 50%, 75%, and 100% patient-matched ascites fluid mixed with advanced RPMI medium supplemented with 10% FBS, 1% penicillin/streptomycin (Sigma Aldrich, MO, USA) and 1% L-glutamine (Sigma Aldrich, MO, USA). Immunofluorescent staining of OCT-embedded spheroids Spheroids were generated from 2000 TYK-nu cells over 72 h in Ascites 1, Ascites 2, and standard cell culture medium as described above. Spheroids were manually washed 3 × with PBS, incubated in 40% sucrose for 16 h, and snap frozen on dry ice in Tissue-Tek optimal cutting temperature (OCT) compound (Sakura, Tokyo, Japan). Following storage at -80° C, 10 µm sections were cut with a Leica CM1850 cryostat (Leica Biosystems, Vic, Australia) and mounted to Superfrost Plus Microscope Slides (Thermo Fisher Scientific, Vic, Australia). Excess OCT was removed with a PBS wash, and sections were fixed and permeabilized with the BD Cytofix/Cytoperm Fixation/Permeabilization kit (BD Biosciences, NSW, Australia), which included fixation for 15 min and washing with permeabilization solution. Sections were blocked with CAS-Block Histochemical Reagent (Thermo Fisher Scientific, Vic, Australia) for 45 min in a humidified chamber and sequentially stained with 1:100 TGM2 recombinant rabbit monoclonal antibody (JU30-02; Thermo Fisher Scientific, Vic, Australia) overnight at 4° C and 1:400 Alexa Fluor™ Plus 555 donkey anti-rabbit IgG secondary antibody (A32794; Thermo Fisher Scientific, Vic, Australia) for 1 h prior to counterstaining with 1:1000 DAPI (Vector Laboratories, CA, USA). Images were acquired with the ZEISS LSM 700 (Carl Zeiss AG, Oberkochen, Germany) using a 10 × objective. Proteomics sample preparation Serum and ascites samples were depleted of high-abundance proteins with the High-Select™ Top 14 Abundant Protein Depletion Spin Columns (Thermo Fisher Scientific, Vic, Australia) following the manufacturer’s instructions. Cell pellets were resuspended on ice in 200 µL RIPA buffer (Merck Millipore, Bayswater, Vic, Australia) supplemented with 1% (v/v) protease inhibitor cocktail (Sigma Aldrich, MO, USA). The lysate was sheared mechanically by passing it five times through a 26 G needle (Terumo Australia Pty Ltd, NSW, Australia), then clarified by centrifugation at 20,000 × g for 30 min at 4 °C. The resulting supernatant was carefully transferred to a fresh tube. Proteins were precipitated overnight at − 20 °C with four-fold ice-cold acetone (ChemSupply, SA, Australia). The protein pellets were collected after centrifugation at 20,000 g for 10 min at − 9 °C, and the supernatant was discarded. Two additional washes with ice-cold acetone preceded centrifugation and removal of residual solvent. Pellets were air-dried on ice for 10 min before proceeding to downstream proteomic workflows. Sample preparation for mass spectrometry analysis followed the protocol set out by Doellinger et al. [33]. Briefly, the cell pellets were resuspended and acidified in trifluoroacetic acid (TFA, Sigma Aldrich, MO, USA), incubated at room temperature for 2 min, and neutralized with 2 M Tris base (Sigma Aldrich, MO, USA) using 10 times the volume of TFA. Reduction and alkylation were carried out by incubating the samples at 95 °C for 5 min with Tris(2-carboxyethyl) phosphine (Merck Life Science, Bayswater, Vic, Australia) to a final concentration of 10 mM and 2-Chloroacetamide (Sigma Aldrich, MO, USA) to a final concentration of 15 mM. After cooling to room temperature, protein concentration was estimated through tryptophan fluorescence [34]. Approximately 100 μg of the proteins from each sample were digested for 20 h at 37 °C with Trypsin Gold (Promega Corporation, WI, USA) at a 1:100 protein-to-enzyme ratio. The digestion was stopped using TFA, and the samples were centrifuged at 20,000 g for 10 min. Peptides were then desalted using Sep-Pak C18 1 cc cartridges (Waters Corporation, NSW, Australia) and vacuum-dried at 50 °C. The dried samples were resuspended in 10 μL of water, and the peptide concentration was estimated using a NanoDrop One/Oneᶜ (Thermo Fisher Scientific, Vic, Australia) at 205 nm. Before LC–MS analysis, samples were adjusted to a peptide concentration of 250 ng/μL ins 0.1% formic acid (Merck Life Science, Bayswater, Vic, Australia). Proteomics data acquisition LC–MS/MS analysis was performed on an EASY-nLC 1200 system (Thermo Fisher Scientific, Vic, Australia) coupled to an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific, Vic, Australia). Approximately 1 μg of peptide sample was loaded at a flow rate of 600 nL/min onto an in-house packed C18 fused silica column (75 μm inner diameter, 360 μm outer diameter, 25 cm length) packed with 1.9 μm ReproSil-Pur 120 C18-AQ particles (Dr. Maisch HPLC GmbH, Ammerbuch, Germany). The column was heated to 50 °C and peptides were separated using a 70 min linear gradient of 3–20% acetonitrile in 0.1% formic acid at 300 nL/min. Two compensation voltages (− 50 and − 70 V) were alternately applied using a FAIMS Pro interface (Thermo Fisher Scientific, Vic, Australia) to adjust the entry of ionized peptides into the mass spectrometer. The mass spectrometer was operated in a data-dependent acquisition manner in positive ion mode. MS scans (m/z 300–1500) were recorded at resolution 60,000 (m/z 200). MS/MS scans were measured at resolution 15,000 after multiply charged peptide precursors (minimum intensity 10,000) were sequentially fragmented with higher energy collision dissociation (HCD) at 27.5% normalized collision energy. A dynamic exclusion period of 40 s was specified, and cycle times were restricted to 1.5 s. Proteomics data processing Raw FAIMS files were processed using FreeStyle 1.8 SP2 software (Thermo Fisher Scientific, Vic, Australia), where data corresponding to individual CV settings (-50 and -70) were exported as separate raw files and subsequently analyzed using MaxQuant software version 1.6.3.4 [35]. The exported raw files from the two CV settings were treated as distinct fractions from the same sample. MS2 spectra from all raw files were searched against the canonical reviewed Homo sapiens UniProtKB (2020_06). Enzyme specificity was set to “Trypsin/P”, the minimal peptide length was set to 7, and the maximum number of missed cleavages was set to 2. A maximum of five modifications per peptide were allowed; the mass tolerance for the first and main search of precursor ions was set to 20 ppm and 5 ppm, respectively, and the mass tolerance for secondary fragment ions was set to 0.02 Da. Carbamidomethylation of cysteine was searched as a fixed modification. “Acetyl (Protein N-term)” and “Oxidation (M)” were set as variable modifications. “Match between runs” was activated, and the MaxLFQ algorithm was used for label-free quantitation. Results were filtered at a false discovery rate (FDR) of 1% at the protein and peptide spectrum match level. Proteomics data analyses Reverse proteins, potential contaminants listed in the cRAP database [https://www.thegpm.org/crap/], and proteins inferred from peptides observed only at the MS1 level were excluded from subsequent analyses. The subcellular localization of each protein was retrieved from UniProtKB and manually annotated for clarity. A two-fold average abundance change and q-value of < 0.05 were required for proteins to be considered candidates. Protein abundance changes induced by ascites across 2D and 3D models were also assessed and visualized in a concordance plot. Proteins were considered highly concordant when fold changes in 2D and 3D dimensionalities were of the same direction and differed by ≤ 0.5 log2-fold units, moderately concordant when the difference was between 0.5 and 1 unit, and discordant when changes were opposite in sign or differed by > 1 unit. An over-representation analysis was conducted on all proteins detected in serum and ascites fluids using the REACTOME pathway database [36], with the top 30 mapped pathways enriched per sample visualized in a clustered heatmap based on false discovery rate (FDR)-adjusted p-values. Gene set enrichment analysis (GSEA) was then performed using REACTOME to compare induced pathways in 2D and 3D cultures of SKOV-3 cells in Ascites 1 and Ascites 2 versus medium, as well as Ascites 1 versus Ascites 2. All heatmaps were generated with Euclidian hierarchical clustering and ‘complete’ linkage methods using the pheatmap R package [37]. The five most significantly enriched biological processes in proteins uniquely expressed in Ascites 1 were identified using DAVID (NIH) [38] and visualized using a custom R-script utilising the ggplot2 package [39]. Principal component analyses were performed and visualized using the EnhancedVolcano R package [40] and upset plots were generated using the UpsetR package [41]. Bar plots were generated using GraphPad Prism 10 [www.graphpad.com]. Results Comparative proteomics of depleted ascites To identify proteins that are specifically enriched in the tumor microenvironment from those commonly found in circulation, we compared the proteomes of Ascites 1–3 sourced from three ovarian cancer patients during cytoreductive surgery prior to adjuvant chemotherapy to healthy serum. The top 14 most abundant proteins were depleted prior to mass spectrometric analysis due to the presence of highly abundant proteins such as albumin and immunoglobulins that suppress less abundant ion signals during analysis. Of the 845 proteins identified across all runs, 793 remained after removal of potential contaminants, reverse proteins, and proteins only detected at the MS1 level. Of these, 298 were detected in all four ascites and serum samples (Fig. 1a) and 67 were uniquely identified in the ascites samples, suggesting potential tumor microenvironment-specific proteins. Interestingly, 111 proteins were detected only in serum. Several proteins were found to be specific to individual ascites samples, including 63 unique to Ascites 1, 95 to Ascites 2, and 54 to Ascites 3, reflecting patient-specific variability. Importantly, a clear distinction was observed between the serum and ascites samples along the first component of the PCA plot, with ascites samples separated by patient origin along the second component (Fig. 1b). Quadruplicate injections yielded highly reproducible data with minor technical variability.Fig. 1Comparative proteome analysis of ascites from three patients and healthy serum from quadruplicate injections. (a) An upset plot showing shared and unique protein groups alongside a plotted (b) principal component analysis which separated ascites from serum along the first component, and ascites from different patients along the second component. Low variability was observed between technical replicates. (c) An over-representation analysis using the thirty-most enriched REACTOME pathways across ascites and serum samples; * = FDR-adjusted pathway enrichment p < 0.05. (d) Relative abundances of secreted, ECM, and membrane proteins that were absent from serum but detected in all ascites, averaged from replicate injections Comparative proteome analysis of ascites from three patients and healthy serum from quadruplicate injections. (a) An upset plot showing shared and unique protein groups alongside a plotted (b) principal component analysis which separated ascites from serum along the first component, and ascites from different patients along the second component. Low variability was observed between technical replicates. (c) An over-representation analysis using the thirty-most enriched REACTOME pathways across ascites and serum samples; * = FDR-adjusted pathway enrichment p < 0.05. (d) Relative abundances of secreted, ECM, and membrane proteins that were absent from serum but detected in all ascites, averaged from replicate injections High-level biological functions were assessed through over-representation analysis of the 30-most enriched REACTOME pathways across biological fluids (Fig. 1c). Expectedly, hemostasis-related terms such as coagulation, platelet activation/signaling/aggregation, hemostasis, heme scavenging, and the clotting cascade, along with the innate immune system, were over-represented in serum versus malignant ascites. Regulation and actioning of the complement cascade, as well as regulation of insulin-like growth factor (IGF) transport and uptake, were similarly significant in both fluid types. Terms relating to protein phosphorylation, neutrophile degranulation, and the extra-cellular matrix (ECM)— in particular, ECM proteoglycans and ECM organization— were markedly higher in all ascites compared to serum. Of note was the distinct profile of Ascites 2, which resided in a separate branch of the ascites clade. To further understand differences between these fluids at the individual protein level, proteins uniquely present in all ascites but not in serum were explored. Of 67 proteins, 19 were either secreted, ECM, or membrane proteins (Fig. 1d), aligning with the significantly enriched terms in ascites relating to the ECM. Several secreted/ECM proteins were uniquely present in ascites samples, including collagens (COL1A2, COL3A1, COL5A1, COL5A2, COL6A2), cell adhesion glycoprotein nidogen-2 (NID2), and laminin 4 (LAMA4). The secreted matrix remodeling associated 5 (MXRA5) and phospholipid transfer (PLTP) proteins were also highly expressed, as were glypican-1 (GPC1) and cadherin 6 (CDH6) membrane proteins. Phenotype of cells grown as three-dimensional cell clusters in ascites and medium Impact of patient-derived ascites on 3D ovarian cancer cell growth To investigate the effect of ascites on tumor cell aggregation and resulting morphology, two ovarian cancer cell lines were cultured in ultra-low attachment plates as 3D spheroids with standard medium, Ascites 1, and Ascites 2, and imaged at 24 and 96 h (Fig. 2a). While spheroid density increased with time in all conditions, more condensed spheroids formed in ascites compared to standard medium, with Ascites 2 having a particularly strong effect. This indicates that the molecular differences between patients’ ascites can alter three-dimensional ovarian cancer cell phenotypes. This effect was replicated in spheroids formed from primary cells cultured in patient-matched ascites (Fig. 2b) and validated in additional cell lines (Fig. 2c). The notion that ascites concentration influences spheroid density was also supported by culturing primary ovarian cancer cells in 0%, 5%, 75%, and 100% patient-matched ascites (Supplementary Figure S1). In summary, growth medium has a direct influence on spheroid formation, showing patient-specific effects of malignant ascites across immortalized and primary cell cultures.Fig. 2Malignant ascites induces phenotypic changes in three-dimensional cell clusters. (a) 2000 TYK-nu and A2780 cells were seeded in each well of 96-well ultra-low adhesive plates and cultured in EMEM medium and two patients’ ascites and imaged at 24 and 96 h to assess spheroid density. (b) 2000 patient-derived cells were cultured in medium or 100% patient-matched ascites and imaged after 72 h. (c) 2000 SKOV-3, COV-362 and A2780 cells were cultured per well in appropriate medium or 100% ascites from two patients’ ascites and imaged after 72 h. All are representative examples of at least three technical replicates imaged with an inverted brightfield digital microscope. (d) The fold-change abundance and significance of proteins observed in spheroid-promoting Ascites 2 versus Ascites 1 displayed in a volcano plot alongside (e) the DAVID biological processes enriched in proteins uniquely identified in Ascites 1 but not Ascites 2 Malignant ascites induces phenotypic changes in three-dimensional cell clusters. (a) 2000 TYK-nu and A2780 cells were seeded in each well of 96-well ultra-low adhesive plates and cultured in EMEM medium and two patients’ ascites and imaged at 24 and 96 h to assess spheroid density. (b) 2000 patient-derived cells were cultured in medium or 100% patient-matched ascites and imaged after 72 h. (c) 2000 SKOV-3, COV-362 and A2780 cells were cultured per well in appropriate medium or 100% ascites from two patients’ ascites and imaged after 72 h. All are representative examples of at least three technical replicates imaged with an inverted brightfield digital microscope. (d) The fold-change abundance and significance of proteins observed in spheroid-promoting Ascites 2 versus Ascites 1 displayed in a volcano plot alongside (e) the DAVID biological processes enriched in proteins uniquely identified in Ascites 1 but not Ascites 2 Given the distinct phenotypic differences of spheroids cultured in Ascites 1 and 2, the proteomes of these fluids were compared to identify potential contributing proteins (Fig. 2d). Protein identification numbers were similar across samples, with an average of 496 proteins identified in each ascites sample and 471 in serum. Basal cell adhesion molecule (BCAM), hornerin (HRNR), and keratins (KRT1/KRT2/KRT5/KRT9/KRT10) were among those more abundant in Ascites 1, whereas heat shock protein 90 isoforms (HSP90AB1, HSP90B1, HSP90AA1), LAMA2, S100A8/S100A9, and insulin growth factor 2 (IGF2) were among the proteins more abundant in Ascites 2. Kallikreins (KLK5/KLK6/KLK10) and other barrier-forming proteins were uniquely present in Ascites 1, whereas periostin (POSTN), growth factors (CTGF, EGFR), and integrins (ITGB2, ITGAM) were uniquely present in Ascites 2. Differences in the proteomes of Ascites 1 and 2 were explored through assessment of enriched biological process (Fig. 2e). Barrier-forming cornification and keratinization pathways were significantly enriched in non-spheroid-promoting Ascites 1, which were largely driven by elevated keratin intermediate filaments. This pattern matches the increased abundances of ECM-cleaving kallikreins and opposes the enrichment of ECM-situated collagens and laminins in spheroid-promoting Ascites 2. Neither cytokines nor periostin induce significant phenotypic changes in 3D-cultured ovarian cancer cells Given the unique presence of POSTN in spheroid-promoting Ascites 2, we assessed whether its inclusion in cell culture medium induces a phenotypic change during SKOV-3 spheroid generation. This was expanded to a panel of selected pro-inflammatory and tumor-associated cytokines, given their well-documented roles in modulating cell proliferation, survival, inflammation, and ovarian cancer progression [42]. After 72 and 144 h, SKOV-3 cells formed spheroids under all conditions, with moderate variation in spheroid sizes and shapes (Fig. 3). These data indicate that neither the addition of cytokines nor POSTN, either alone or in combination, were sufficient to induce a change in phenotype as it has been observed when the cells were seeded in Ascites 2. None of the conditions altered viability (data not shown).Fig. 3Cytokines and POSTN induce minimal phenotypic changes of SKOV-3 cells grown in 3D. SKOV-3 cells were grown in medium, medium supplemented with cytokines (0.25 ng/ml IL1-β, 1 ng/ml IL6, 10 ng/ml TNFα, 10 ng/ml TGFβ, medium supplemented with 50 ng/well POSTN or 100% Ascites 2. After 72 h and 144 h the cells were imaged using an inverted microscope. One representative well of cells seeded in triplicates in a 96-well plate is shown per condition Cytokines and POSTN induce minimal phenotypic changes of SKOV-3 cells grown in 3D. SKOV-3 cells were grown in medium, medium supplemented with cytokines (0.25 ng/ml IL1-β, 1 ng/ml IL6, 10 ng/ml TNFα, 10 ng/ml TGFβ, medium supplemented with 50 ng/well POSTN or 100% Ascites 2. After 72 h and 144 h the cells were imaged using an inverted microscope. One representative well of cells seeded in triplicates in a 96-well plate is shown per condition To assess whether the phenotypic changes induced by malignant ascites altered chemotherapy response, primary ascitic cells were cultured as spheroids in increasing concentrations of patient-matched ascites. Larger and more compact spheroids formed in the presence of ascites (data not shown), and the viabilities of spheroids cultured in elevated ascites concentrations were higher than those incubated in standard cell culture medium (Fig. 4a). However, the normalized percentage of cell death following carboplatin treatment did not differ significantly between culture conditions (Fig. 4b), suggesting that phenotypic changes and increased baseline viability are not predictive of drug response.Fig. 4Ex vivo carboplatin (CBP) treatment caused similar relative decreases in spheroid viability regardless of ascites content in cell culture medium, despite altered baseline viability. (a) Primary ascites cells were cultured as spheroids in four concentrations of malignant ascites added to RPMI cell culture medium, with baseline viabilities normalized to the medium-only control. (b) The change in spheroid viability following treatment with 300 μM CBP across four culture conditions. Significance calculated with Tukey-corrected t-tests from at least three replicates; **p < 0.005, ***p < 0.0005 Ex vivo carboplatin (CBP) treatment caused similar relative decreases in spheroid viability regardless of ascites content in cell culture medium, despite altered baseline viability. (a) Primary ascites cells were cultured as spheroids in four concentrations of malignant ascites added to RPMI cell culture medium, with baseline viabilities normalized to the medium-only control. (b) The change in spheroid viability following treatment with 300 μM CBP across four culture conditions. Significance calculated with Tukey-corrected t-tests from at least three replicates; **p < 0.005, ***p < 0.0005 Proteomic analysis of 2D and 3D SKOV-3 cultures grown in patient ascites Given the lack of effect of interleukins and POSTN on the phenotype of three-dimensional immortalized cell cultures, we aimed to understand how factors contained within ascites affect ovarian cancer cell proteomes. To do this, SKOV-3 cells were grown as spheroids in standard DMEM medium, Ascites 1, or Ascites 2, and their proteomes were compared. Of the 6,142 proteins detected across all samples, 6074 remained following the removal of potential contaminants, reverse, and proteins detected only at MS1 level (Supplementary Figure S2a). Of these, 3234 were identified in all conditions, with 968 proteins uniquely identified in 3D SKOV-3 cells grown in ascites of any patient origin and 165 commonly identified in spheroids cultured in both ascites. Interestingly, 1174 proteins were uniquely identified in spheroids grown in standard medium, with differences reflected in the segregation of all samples (Supplementary Figure S2b). However, hierarchical clustering grouped spheroids that were cultured in Ascites 2 separately from spheroids grown in medium or Ascites 1 (Fig. 5a). Incubation with ascites therefore leads to patient-specific changes in the proteome of ovarian cancer cells, highlighting the impact of the tumor microenvironment on spheroid biology.Fig. 5Comparative proteomic analysis of SKOV-3 cells grown as 2D adherent 3D tumor spheroids and 2D adherent in Ascites 1, Ascites 2, or standard cell culture medium, displayed as (a) a global proteome heatmap and (b) an upset plot. (c) The concordance of protein expression changes between 2 and 3D dimensionalities when cultured in ascites versus medium. Expression within Ascites 1 and Ascites 2 were averaged from all replicates for clarity. (d) Uniquely present and upregulated proteins in adherent SKOV-3 cells that were cultured in medium and ascites, annotated with the number which are classified as secreted, or reside in the ECM or membrane. Comparative proteomic analysis of SKOV-3 cells grown as 2D adherent 3D tumor spheroids and 2D adherent in Ascites 1, Ascites 2, or standard cell culture medium, displayed as (a) a global proteome heatmap and (b) an upset plot. (c) The concordance of protein expression changes between 2 and 3D dimensionalities when cultured in ascites versus medium. Expression within Ascites 1 and Ascites 2 were averaged from all replicates for clarity. (d) Uniquely present and upregulated proteins in adherent SKOV-3 cells that were cultured in medium and ascites, annotated with the number which are classified as secreted, or reside in the ECM or membrane. The abundances of secreted, ECM, and membrane proteins in SKOV-3 spheroids were compared between those grown in medium and those grown in ascites from two separate patients and are listed in Supplementary Table S2. Collagens (1A1, 1A2, 3A1, 5A1, 5A2, 6A2) and apolipoproteins (A-I, A-IV, C-IV, D, E), were significantly more abundant in spheroids cultured in patient ascites, with many not detected in samples cultured in standard cell culture medium. Coagulation factors (X, XI, XII), haptoglobin, hemopexin, alpha-1B-glycoprotein, and ceruloplasmin are plasma proteins that were detected only in ascites-cultured spheroids. Given that malignant ascites originates from increased capillary permeability, such proteins may simply be derived from the ascites, rather than produced by spheroids in response to their environment. To confirm, we repeated the experiment using adherent cultures that can be processed more vigorously to reduce contamination of proteins that may stick to surfaces and cells. In this setup, SKOV-3 cells were grown in monolayer using either standard medium or 100% ascites from two patients. Hierarchical clustering robustly separated samples by growth dimensionality, as expected, and further distinguished Ascites 2–cultured spheroids from those grown in standard medium or Ascites 1 (Fig. 5a). This consistent segregation indicates that Ascites 2 exerts a strong, condition-independent effect on the cellular proteome. While a core set of proteins (35%) was detected across all conditions, 188 proteins were completely undetectable in all spheroidal cultures (Fig. 5b), indicating a subset of proteins which might be associated specifically with adherent growth. To assess how similarly 2D and 3D cultures respond to ascites, we compared the change in each protein’s abundance induced by ascites in 2D versus 3D and classified proteins according to their agreement in both direction and magnitude (Fig. 5c). Following verification that neither Ascites 1 nor Ascites 2 produced significantly different magnitudes of protein abundance shift (data not shown), we identified 1,265 (45%) proteins in all conditions with highly concordant responses to ascites across dimensionalities, whereas 1,137 (40%) exhibited discordant changes between 2D and 3D cultures. The pathways induced by ascites in 2D and 3D cultures were next assessed by GSEA. Besides enriched complement and coagulation cascade, which could be contamination, ascites exposure consistently resulted in enriched plasma lipoprotein assembly and remodeling pathways (Supplementary Figure S3). The strongest signals were observed in 3D cultures treated with Ascites 2, indicating a shared proinflammatory and metabolically altered milieu. Integrin-related signaling, scavenger receptor–mediated uptake, and collagen-related pathways showed moderate but reproducible enrichment, supporting a shift toward enhanced matrix remodeling in ascites-cultured cells. Direct comparison of pathways enriched by Ascites 1 and Ascites 2 revealed more subtle but concordant differences centered on translational control, amino-acid metabolism, and a small subset of guidance and cytoskeletal pathways, including collagen trimerization, consistent with patient-specific tuning of ECM-handling and stress-response programs (Supplementary Figure S4). The ECM-relevant pathway-level changes and the representation of secreted, ECM, and membrane proteins in cell-free ascites fluid provided a mechanistic rationale for the marked increase in such proteins within ascites-treated cultures. Combined with spheroid formation, we assessed the number of such proteins that were upregulated in either culture condition (Fig. 5d). The vast majority (155) were upregulated or uniquely present in cells grown in ascites, with 151 being uniquely identified in this culture condition. Of these proteins, 22 were also more abundant in 3D cultures compared to those cultured in standard medium (Fig. 6a). Twenty-one were only identified in adherent cells cultured in ascites, with apolipoproteins, collagens, coagulation factors, and cadherin 6 commonly detected in 2D and 3D ascites cultures (Fig. 6a). Interestingly, TGM2 was the only protein of this group to be detected in all medium- and ascites-cultured spheroidal and adherent cultures, and there was a significantly increased level detected in spheroids cultured in ascites compared to medium. To confirm that incubation with ascites induces TGM2 expression, IF staining of TGM2 was conducted in spheroids grown in medium, Ascites 1 and 2 (Fig. 6b). TGM2 was only detected in ascites-cultured spheroids, aligning with mass spectrometry data.Fig. 6ECM, secreted, and membrane proteins that were upregulated in both 2D and 3D SKOV-3 cells cultured in Ascites 1 and Ascites 2 versus medium revealed ascites-induced TGM2 expression. (a) Clustered heatmap displaying the relative abundances of ECM, secreted, and membrane proteins in 2D and 3D cultures, identified through mass spectrometry. An asterisk denotes TGM2, which was the only listed protein that was identified in ascites and medium culture conditions. (b) IF staining of spheroids generated from TYK-nu immortalized ovarian cancer cells in EMEM medium and two patients’ malignant ascites, validating the induction of TGM2 by ascites in spheroids. One representative spheroid from each condition is shown ECM, secreted, and membrane proteins that were upregulated in both 2D and 3D SKOV-3 cells cultured in Ascites 1 and Ascites 2 versus medium revealed ascites-induced TGM2 expression. (a) Clustered heatmap displaying the relative abundances of ECM, secreted, and membrane proteins in 2D and 3D cultures, identified through mass spectrometry. An asterisk denotes TGM2, which was the only listed protein that was identified in ascites and medium culture conditions. (b) IF staining of spheroids generated from TYK-nu immortalized ovarian cancer cells in EMEM medium and two patients’ malignant ascites, validating the induction of TGM2 by ascites in spheroids. One representative spheroid from each condition is shown Discussion Malignant ascites is a key driver of metastatic spread and chemoresistance in ovarian cancer, in part through the formation of multicellular spheroids that resist anoikis and chemotherapy. By integrating proteome profiling of patient-derived ascites, 2D and 3D SKOV-3 models, and ex vivo drug response assays, our study shows that: (i) ascites is enriched for ECM and secreted proteins that promote spheroid aggregation; (ii) ascites from different patients, particularly Ascites 2, impose non-overlapping, dimensionality-dependent proteomic signatures on tumor cells; and that (iii) despite these differences, standard cell culture medium can still support robust platinum dose–response testing, which yields practical implications for functional precision medicine approaches using ex vivo cultures. First, we compared the proteomes of cell-free ascites from three patients to healthy serum. The strong separation of serum from all ascites samples in the hierarchical clustering and the enrichment of immune system and coagulation pathways reflects the distinct biological compartmentalization of ascites from wider circulation. The over-representation of ECM proteins, which provide structural stability and regulate key cancer-related processes [21], was exemplified by increased abundances of several collagens. Implicated in cancer cell survival, the degradation of collagen 1 into glutamine for resorption by tumor cells has been shown to satisfy their high metabolic demand [43]. Collagen density, organization, and signaling also promote metastasis [44] and chemotherapy resistance [45], potentially through increased ECM stiffness that modulates drug efflux [46]. Importantly for ovarian cancer progression, increased ECM density promotes larger spheroid clusters [47]. The unique presence of other ECM proteins in ascites that are also implicated in disease progression include cadherin-6 [48], laminin subunit alpha-4 [49], and fibulin 2 [50]; however, their complex and often contrasting roles have prevented their use as molecular markers of disease. The detection of several ECM-regulating annexins (ANXA1, ANXA2, ANXA5) solely in ascites supports the proposed function of this protein family as molecular markers of ovarian cancer. However, the direction of change is often inconsistent between studies, with annexins having been reported as markers of cisplatin [51] and paclitaxel [52] resistance. For instance, ANXA2 overexpression has been observed in ovarian cancer tissues, and its upregulation correlates with worsened overall survival and ascites. Proteins uniquely detected in ascites samples include laminins and SPON1, which have been implicated in disease progression and metastasis [53, 54]. Importantly, the present study demonstrates ascites to be amenable to the detection of ECM and secreted proteins and their known regulators through mass spectrometry. Differences in ascites proteomes were demonstrated in the significant upregulation of several heat shock 90 isoforms (e.g., HSP90AB1, HSP90B1, and HSP90AA1) in the spheroid-promoting Ascites 2, along with S100A8/S100A9 [55] and integrins (e.g., ITGB1) [56], which have been reported as potential negative regulators of anoikis. Proteins that were uniquely present in spheroid-promoting Ascites 2 but not Ascites 1 included POSTN, growth factors (CTGF, EGFR) and integrins (ITGB2, ITGAM). CTGF is a well-established downstream mediator of TGF-β that drives COL1A1 and broader collagen I deposition in fibrotic and tumor settings, which has been shown to support tightly compacted, drug-resistant tumor spheroids [57]. These data suggest that Ascites 2 may provide a CTGF-, integrin- and EGFR-competent ECM that favors collagen I–dependent cell–to-cell and cell–to-matrix adhesion, consistent with the more compact spheroid morphologies observed in both cell lines and primary ascites cells. A large subset of proteins only identified in Ascites 1 instead promote barrier and cornified envelope formation (e.g., FLG/FLG2, KLK5/KLK6/KLK10, and XP32), which typically stiffens stratified epithelia [58]. In this context, such proteins are likely to reduce motility and spheroid formation. Such patient-specific pathways were reflected in the morphological and viability changes induced in cell line spheroids. The elevation of spheroid viability with increasing malignant ascites content in cell culture medium supports the known role of malignant ascites in disease progression and dissemination [59]. However, the use of standard cell culture medium in ex vivo treatment response assessments remains appropriate, given the similar proportional decreases in viability observed following treatment in both conditions with carboplatin, the most common platinum-based ovarian cancer chemotherapeutic. We hypothesized that a bidirectional relationship exists between the patient-dependent aggregation capabilities of ascites alongside standard cell culture medium and the proteomes of cultured spheroids. We used mass spectrometry to profile the proteomes of SKOV-3 cells that were cultured as three-dimensional spheroids in ascites versus cell culture medium, which emulates the microenvironment present in malignant ascites. Interestingly, we observed separate hierarchical clustering of spheroids grown in Ascites 2 from those grown in both Ascites 1 and standard cell culture medium. Spheroids cultured in Ascites 2 contained significantly fewer proteins than any other culture condition, with 567 proteins that were present in spheroids cultured in Ascites 1 or medium not identified. Secreted and ECM proteins produced by tumor or stromal cells mediate cell-to-cell communication, tumor growth, and metastasis [60] and therefore were the focus of our analyses. The abundance of apolipoproteins in ascites-cultured SKOV-3 spheroids, which regulate lipid metabolism, transport, and enzyme activity in cancer [61] reflect a complex microenvironment. The elevated or unique abundance of collagens in ascites-cultured spheroids reflects their previously discussed reported roles in spheroid formation and chemotherapy response. A previous analysis of malignant ascites proteomes from gastric cancer patients identified POSTN, a secreted protein, as one of four protein candidate markers of gastric cancer [62]. While not significantly more abundant in SKOV-3 spheroids incubated in Ascites 1, POSTN was significantly more abundant in Ascites 2-cultured spheroids compared to those formed in standard culture medium. Given that Ascites 2 drove spheroid aggregation and condensation, and that POSTN expression is linked to ovarian tumor invasion, dissemination [63], and recurrence [64], we hypothesized that POSTN— and chemokines also known to create a feedback loop that enhances cancer cell proliferation and invasiveness [65]— would alter spheroid formation of SKOV-3 cells. We instead observed that the separate addition of these factors did not induce morphological changes during the chosen timeframe. Given the complex role of these factors in vivo, further studies could assess whether their addition may alter signaling pathways at the transcriptome or proteome levels. SERPINA1, a member of the serine proteinase inhibitor protein family, was also uniquely detected in the spheroid-promoting Ascites 2 but absent from Ascites 1. However, it was also detected in SKOV-3 spheroids cultured in Ascites 1, implying that incubation with ascites can induce SERPINA1 expression within cancer cell lines. Interestingly, our previous studies have demonstrated SERPIN A protein abundance to be significantly higher in spheroids collected directly from the ascites of chemotherapy sensitive patients, opposing elevated SERPIN B abundances in spheroids of resistant patients [21]. SERPINA10 has also been associated with improved progression-free and overall survival in the publicly available TCGA cohort, which was supported by further IHC-based validation in a patient cohort using fresh-frozen tissue lysates [66]. A reverse trend has been seen in serum, where SERPINA1 expression correlated with platinum resistance [67]. To ensure that the differences in the proteomes observed have not been caused by ascitic proteins sticking non-specifically to labware and cells, we performed subsequent proteome analyses of adherent SKOV-3 cells collected after washing. In line with previous comparisons of 2D and 3D cell cultures [68–70], distinct proteome differences were observed between spheroid and adherent cell cultures. Previous analysis of the effects of ovarian cancer ascites on adherent SKOV-3 cells visualized 300 protein spots on a two-dimensional gel and identified 20 proteins spots that were different. Upregulated proteins included PDZ domain-containing protein, GTP-binding nuclear protein Ran, zinc finger protein 268, collagen alpha 6 (VI) chain, actin B, keratin type II cytoskeletal 8, Synaptotagami-like protein 3 and vimentin [71]; however, none were differentially abundant in our analyses of 2D or 3D cells cultured in ascites. Changes to the global proteomes of 2D and 3D cells were relatively concordant when cultured in ascites compared to standard medium. The large fraction of proteins with highly concordant responses to ascites likely represents a core, dimensionality-independent program; however, no biological pathway was preferentially altered among concordant or discordant proteins. Pathway analyses did support and extend protein-level observations. The prominent enrichment of complement and coagulation cascades in ascites-treated cultures, together with platelet and lipoprotein pathways, is consistent with prior reports that ovarian cancer ascites constitutes a highly active but ultimately tumor-promoting complement/immune context rather than a purely cytotoxic effector compartment [72, 73], with particular effects observed in ascitic extracellular vesicles [9]. The concomitant enrichment of integrin-linked signaling, scavenger receptor–mediated uptake, and collagen-related pathways points to ascites reinforcing matrix engagement and ECM remodeling programs, in line with studies showing that ascites induces β1-integrin expression [74] and that ascites-derived spheroids adhere to, and migrate on, peritoneal ECM via β1-integrins and fibrin-rich matrices [75]. Notably, the concordant but subtle differences between Ascites 1 and Ascites 2 in translational control, amino-acid metabolism, and selected guidance/cytoskeletal pathways align with their patient-specific ECM and secreted protein signatures and divergent spheroid morphologies, and mirror reports that ascites microenvironments variably condition pre-metastatic niches and extracellular vesicle–mediated ECM remodeling [9]. Together, these data suggest that ascites is complement-rich and ECM-modifying, providing a plausible mechanistic basis for the observed ECM-centric proteome changes and spheroid morphologies. The observed induction of TGM2 in response to ascites in 2D and 3D ovarian cancer models, and subsequent orthogonal validation in spheroids through IF, warrants further interrogation of its role in ovarian cancer biology. Originally described as a transamidase protein that resides in the ECM, TGM2 is a GTP-binding protein that contributes to cancer progression through signal transduction, ECM modulation, and apoptosis regulation [76], potentially through antagonistic roles [77]. TGM2-deficient models have implicated this protein in spheroid formation and ECM assembly, with fewer and smaller spheroids coinciding with elevated apoptosis [78], which has been reported to relate to reduced MEK1/2 and ERK1/2 activity [79] and stimulation from EGF that increases cell contractility [80]. Knockdown and inhibition of TGM2 have therefore reduced spheroid formation in epidermal squamous cell carcinoma [81], with abundance impacting chemotherapy response in cisplatin-resistant ovarian cancer spheroids [82]. Given these findings, studies of its induction in ovarian cancer ascites are necessary to reveal downstream implications on disease progression, particularly in the context of opposing activity in apoptosis [77]. Conclusions The present study demonstrated that malignant ascites supports the formation of spheroids in ovarian cancer cells. Distinct morphological changes were observed in both cell line and ascites-derived spheroids in the presence of ascites, including increased compactness and aggregation. These phenotypic changes were reflected at the proteome level, at which the abundances of proteins such as collagens, laminins, and annexin A2 were altered. In addition, ascites induced expression of TGM2— a protein with known roles in treatment resistance, metastasis, and spheroid formation— in 2D and 3D SKOV-3 cell cultures. Orthogonal IF staining of spheroids supported the observed induction of TGM2 expression by ascites. Finally, our findings demonstrate that standard cell culture medium can be used in ex vivo chemotherapy response assessments despite reduced baseline viability of spheroids compared to growth in patient-matched ascites. The small number of ascites samples limits the generalizability of these data, especially in the context of the pronounced phenotypic and proteome differences observed between malignant ascites samples. It would be prudent to include patients with varying clinical treatment responses among a larger cohort to validate and extend upon these findings. Supplementary Information Additional file 1. Additional file 2. Additional file 3. Additional file 1. Additional file 2. Additional file 3.