Benchmarking Ultra‐Low Attachment and Photopatterned GelMA 3D Culture Platforms for Modeling Cancer Stemness in High‐Grade Serous Ovarian Cancer

ABSTRACT

High‐grade serous ovarian cancer (HGSOC) exhibits pronounced intratumoral heterogeneity and chemoresistance, demanding experimental platforms that more accurately recapitulate stemness‐driven relapse. Despite increasing use of three‐dimensional (3D) systems in cancer research, their comparative translational performance in ovarian cancer remains unclear. Here, we systematically benchmark two scalable 3D platforms, scaffold‐free ultra‐low attachment (ULA) spheroids and extracellular matrix mimetic photopatterned gelatin methacryloyl (GelMA) hydrogels, using HGSOC cell lines and patient‐derived primary samples. Under matched conditions, both systems enhanced cancer stem cell (CSC) associated phenotypes relative to 2D culture, including induction of stemness markers, CSC surface populations, epithelial to mesenchymal transition signatures, and drug resistance. ULA enabled long‐term spheroid maturation and robust expansion of intrinsically stem‐like populations, whereas GelMA photopatterning provided spatially defined microenvironmental control that promoted rapid and uniform spheroid formation and efficiently induced stem‐like traits in low‐stemness models. Notably, baseline SOX2 expression partially correlated with chemoresistance, supporting the biological relevance of CSC enriched platforms. Collectively, this platform level comparison establishes complementary, biotechnology driven strategies for reproducible modeling of HGSOC stemness and therapy resistance, with implications for translational drug evaluation and precision oncology.

Graphical Abstract and Lay Summary

Ultra‐low attachment (ULA) and photopatterned GelMA were benchmarked in HGSOC cell lines and patient‐derived cultures. Both platforms enhanced CSC‐ and EMT‐associated phenotypes but showed distinct strengths. In OVCAR‐3, ULA supported long‐term spheroid maturation, scalable recovery, and multidrug‐resistant phenotypes, whereas GelMA enabled rapid spheroid formation and enhanced CSC‐associated traits in weakly aggregating models. Baseline SOX2 partially corresponded with drug resistance in patient samples, supporting context‐dependent selection of complementary 3D platforms.

1. Introduction

Ovarian cancer (OC) remains a major clinical challenge and a leading cause of gynecologic cancer–related mortality worldwide, ranking as the eighth most common cancer and cause of cancer‐related death in women, with an estimated 324,000 new cases and 207,000 deaths annually (World Ovarian Cancer Coalition; World Cancer Research Fund International). Despite advances in surgical management and systemic therapies, long‐term outcomes remain poor, with a global five‐year survival rate of 50.9% reported between 2014 and 2020 according to the National Cancer Institute (NIH). High‐grade serous ovarian cancer (HGSOC) is the predominant histological subtype and is initially responsive to standard‐of‐care treatment, with approximately 85% of patients achieving clinical remission following cytoreductive surgery and platinum‐based chemotherapy. However, a substantial fraction of patients exhibit intrinsic resistance (15%–20%) or develop acquired resistance within 6 months (∼25%), ultimately leading to disease recurrence and treatment failure [1, 2]. These features underscore the need for experimental models that more accurately capture the biological complexity and therapy‐resistant states characteristic of HGSOC.

Cancer stem cells (CSCs), also referred to as tumor‐initiating cells (TICs), were first described in acute myeloid leukemia (AML) [3] and have since been implicated as key drivers of tumor heterogeneity, recurrence, and therapeutic resistance across multiple cancer types. The CSC model proposes that a subpopulation of tumor cells possesses self‐renewal capacity, the ability to transition between quiescent and proliferative states, and phenotypic plasticity enabling epithelial–mesenchymal transitions. These properties confer a selective advantage under therapeutic pressure and are associated with poor clinical outcomes [4]. In HGSOC, CSCs are considered major contributors to both intrinsic and acquired resistance [5, 6]. Their ability to persist in a quiescent (G0) state allows evasion of cytotoxic therapies, followed by reactivation and tumor regeneration, thereby promoting disease relapse [7, 8]. Consequently, improving the experimental modeling of CSC‐associated phenotypes is particularly relevant for understanding HGSOC recurrence and resistance mechanisms.

Three‐dimensional (3D) culture systems have emerged as important biotechnological tools for modeling CSC biology and tumor heterogeneity in vitro [9]. While 3D approaches are increasingly well established in several cancer types, including breast, colorectal, and pancreatic cancers, their application and systematic evaluation in ovarian cancer remain comparatively limited. A comprehensive overview of ovarian cancer 3D models was provided in 2022 [10], and more recent work has highlighted the role of 3D culture in chemoresistance and metastatic behavior [11]. Among scaffold‐free approaches, ultra‐low attachment (ULA) plates are widely used to generate spheroids by preventing cell–substrate adhesion. Depending on plate geometry, ULA systems can support either heterogeneous multicellular aggregates or more uniform spheroids [10], offering advantages in reproducibility and throughput compared to earlier methods such as agar or agarose coating [12, 13, 14]. Hanging drop cultures remain a useful alternative but are less compatible with scalable and higher‐yield applications.

Scaffold‐based strategies provide complementary solutions by introducing extracellular matrix (ECM)‐derived cues that more closely approximate the tumor microenvironment [15]. Gelatin methacryloyl (GelMA) is a semi‐synthetic, biocompatible hydrogel derived from gelatin and is widely used in 3D cell culture due to its tunable mechanical properties and chemical stability [16]. GelMA has been employed for spheroid delivery in vivo [17], and offers advantages including matrix metalloproteinase responsiveness, inherent RGD motifs, batch‐to‐batch consistency, and compatibility with photopatterning approaches using masks or molds [18, 19, 20]. In parallel, organoid cultures embedded in basement membrane extracts such as Matrigel are increasingly used to recapitulate tumor architecture; however, their widespread adoption is limited by batch variability, cost, and challenges in scalability [21]. As a result, ULA‐based spheroid cultures and photopatterned GelMA hydrogels [22] remain highly relevant and complementary platforms for translational ovarian cancer research.

Traditional two‐dimensional (2D) culture systems, including the NCI‐60 panel established in the 1980s and later expanded by the Japanese Foundation of Cancer Research [23], have been instrumental in drug discovery and molecular characterization [24, 25, 26, 27]. Nevertheless, extensive comparative studies have demonstrated the limitations of 2D monolayers in modeling tumor heterogeneity and therapy resistance [28, 29, 30, 31]. Both 2D and 3D ex vivo systems approximate in vivo conditions more effectively than monolayer cultures alone [32, 33], yet 2D conditions tend to underrepresent ovarian cancer stem‐like populations. Proliferation‐favoring environments and repeated passaging selectively enrich epithelial phenotypes while reducing stem‐like diversity. In contrast, 3D culture systems consistently enhance CSC‐associated traits and drug resistance [34, 35, 36, 37]. This enrichment may arise from selective survival of CSC‐like cells [38], as well as ECM‐mediated signaling mechanisms that actively promote stemness [39, 40, 41].

Developing CSC‐enriched, clinically relevant 3D models is therefore critical for advancing ovarian cancer biotechnology and translational research. In this study, we perform a focused, platform‐level comparison of two widely used 3D culture systems—ULA‐based spheroid culture and photopatterned GelMA hydrogels—using HGSOC cell lines (OVSAHO and OVCAR‐3) and patient‐derived primary samples. We evaluate CSC‐associated surface markers (CD133 [42], CXCR4 [43], CD117 [44], CD44 [45], all previously reported as ovarian CSC markers [46, 47]), stemness‐related transcription factors (SOX‐2, OCT‐4, NANOG, ALDH1A1 [48], c‐MYC), epithelial–mesenchymal transition (EMT) markers, spheroid morphology and dynamics, and drug response profiles. By benchmarking biological performance alongside practical considerations such as scalability and robustness, we aim to define complementary 3D culture strategies that improve modeling of ovarian cancer stemness, recurrence, and therapeutic resistance.

We hypothesized that scaffold‐free ULA and ECM‐based photopatterned GelMA platforms would differentially enrich CSC‐associated phenotypes in HGSOC models through distinct biological and biophysical mechanisms, and that systematic benchmarking of these complementary systems would clarify their respective advantages for modeling ovarian cancer stemness, therapeutic resistance, and translational applications.

2. Methods

This study was designed as an experimental in vitro investigation comparing two three‐dimensional (3D) culture systems—ULA plates and GelMA hydrogels fabricated as photopatterned constructs using a photomasking‐based UV crosslinking approach (GelMA‐Pm), for their ability to enrich ovarian cancer stem cell (OCSC) features in HGSOC models. The study included established HGSOC cell lines (OVSAHO and OVCAR‐3) and patient‐derived primary ovarian cancer cells. All experiments were conducted at Koç University Research Center for Translational Medicine (KUTTAM), Istanbul, Türkiye, between 2019 and 2024 under standardized laboratory conditions.

2.1. Cell Culture

2.1.1. 2D Conventional Culturing of Cell Lines

OVCAR‐3 and OVSAHO cell lines were obtained from ATCC, cultured with DMEM/F12, 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, 1% L‐glutamine additives, in normal attachment flasks. Cells were passaged at 70% confluency according to proliferation rate. Cell lines were routinely tested for mycoplasma.

2.1.2. 3D Culturing Using ULA Plates

All types of cells were trypsinized from T‐75 flasks (0.25% Trypsin, EDTA) and reduced to single cells through pipetting and a number of 1 × 10⁵–3 × 10⁵ cells were seeded per well into ULA plates (6‐well, Corning), in 3–6 mL of spheroid enrichment medium of DMEM/F12 (Multicell), 0.5% FBS, 1% penicillin/streptomycin, 1% L‐glutamine, 10 ng/mL FGF, 10 ng/mL EGF, 5 µg/mL insulin, cultured in incubators with 5% CO₂ at 37°C. Cells were kept in the incubator for the first 14 days where aggregates start to form. Medium change by adding 3–5 mL spheroid enrichment medium onto the old medium and aggregates was performed at this time point. During the 14–28 days of culture, stable spheroids and stemness‐enriched populations form. Cells can be passaged, optionally reducing to single cells with Accutase and trypsin (0.05 % trypsin, EDTA) solutions and P2 cells will form spheroids in a shorter time and larger in size, starting to form highly compartmentalized loose spheroids. Cells can be passaged up to 3 times after spheroid formation and medium changes can be applied in between. Passaging is applied by a weaker centrifugation than single cells require, at 200 × g for 2 mins, and supernatant is removed, removing much of the dead cell debris but still pelleting the heavy cell clumps and spheroids. Re‐seeding or Accutase application can later be applied to the pellet.

2.1.3. 3D Culturing Using Gelatin‐Methacrylate and Photomasks

GelMA (ZetaMatrix) was synthesized using a one‐pot method, according to the process described in previous research [49]. Briefly, a 10% (w/v) gelatin solution was generated by dissolving 10 g of bovine skin gelatin in 100 mL of carbonate‐bicarbonate buffer (0.25 M, pH 9) at 50°C. Methacrylic anhydride (MA) was gradually added dropwise while simultaneously stirring magnetically at 50°C. To terminate the reaction, the solution's pH was adjusted to 7.4. Subsequently, the solution was dialyzed against distilled water for 7 days at 40°C using a dialysis membrane to eliminate unreacted MA groups and salts. This process was followed by lyophilization and storage at −20°C.

The GelMA (Zetamatrix BV) prepolymer solution was composed by dissolving lyophilized GelMA in DPBS that included the photoinitiator (PI) 2‐hydroxy‐4′‐(2‐hydroxyethoxy)‐2‐methylpropiophenone (Irgacure 2959). The PI solution was first prepared by dissolving Irgacure 2959 powder in the specified volume of DPBS at 80°C, achieving a concentration of 1% (w/v). Next, the lyophilized GelMA was incorporated into the Irgacure 2959 solution to obtain a final concentration of 5% (w/v) GelMA prepolymer solution. The mixture was kept at 37°C and periodically vortexed to ensure the thorough and uniform dissolution of both components.

For cell encapsulation, primary ovarian cancer cells or ovarian cancer cell lines were trypsinized with 0.25% trypsin/EDTA, counted, and the required number of cells were pelleted by centrifugation at 300 × g for 5 min. The pellet was then resuspended in the appropriate volume of 5% (w/v) GelMA prepolymer solution to achieve a concentration of 4 million cells/mL. Hundred microliters of the GelMA pre‐polymer solution containing cells were placed on a spacer with a depth of 0.3 mm, and a glass slide was inverted and placed on top of the deposited pre‐polymer droplet.

Following this, a photomask (designed using CorelDRAW software, printed by Corel Corp., Canada, and A Print, Izmir) was placed on the glass slide and exposed to UV light (with a power density of 6.25 𝑊/𝑐𝑚²) for 50 s using the OmniCure S2000 UV 12 Light Curing System (Excelitas Technologies Corp., Waltham, MA, USA).

Finally, the structures were washed with DPBS to remove non‐crosslinked excess GelMA pre‐polymer and placed in glass bottom culture dishes for 3D cell culture experiments. Spheroid media depicted above was added to the dish. The hydrogel‐coated 3D cell culture frames were maintained in an incubator at 37°C with 5% CO₂ for 14 days. Structures were imaged with a brightfield microscope regularly to track spheroid formation.

2.1.4. Epithelial Cell Isolation From Primary Tumor Tissues and Ascites

Tumor tissue samples were obtained from patients with HGSOC during surgery and placed into sterile containers either filled with ice‐cold 1X DPBS or 1X Hanks’ Balanced Salt Solution (HBSS). Samples were then transferred onto a glass petri dish, washed several times with 1X DPBS and cut into small pieces by using a sterile forceps and a scalpel. The small fragments of tissues were enzymatically digested with 3 mg/mL (2.4U/mL dispase, 0.3 U/mL collagenase) Collagenase/Dispase (Roche Diagnostics) in 1X DPBS and agitated at 37°C for 1 h. After agitation, the cell suspension was transferred onto a 100 µm cell strainer and cells were allowed to pass through the filter to remove any undissociated tissue. Approximately 10 mL of 1X DPBS (prewarmed) was added onto the cell strainer to dilute enzyme solution. The obtained cell suspension was centrifuged at 400 × g for 5 min. Cell pellet was resuspended in 10 mL of DMEM/F12 containing 20% FBS, %1 P/S, and %0.1 Amp B and cultured into a 100 mm petri dish in a 37°C incubator with 5% CO₂. Medium was refreshed whenever needed but not earlier than 4–5 days. Once the primary cells reached 80% confluency, the media was replaced with DMEM/F12 containing 10% FBS, %1 P/S, and %0.1 Amp B and the cultures of primary cells obtained from tumor tissues were ready for subculturing for further experiments.

Besides tumor tissues, ascites samples from HGSOC patients who underwent cytoreductive surgery were aspirated directly into a sterile suction bottle during surgery and used to establish primary ascites cell cultures. After collecting ascites fluid, under sterile conditions, it was split into a number of 50 mL falcon tubes based on the total volume obtained. The samples were centrifuged at 500 × g for 5 min, and the supernatants were discarded. The cell pellets were resuspended in 30 mL of prewarmed 1X HBSS by first adding 27 mL of ddH₂O, followed by 3 mL of 10X HBSS. The gentle pipetting was applied to mix. The centrifugation and resuspension steps were repeated up to three times to remove cellular debris and erythrocytes. Finally, the cleaned cell pellets were resuspended in 7 mL of DMEM/F12 complete medium containing 20% FBS, 1% P/S, and 0.1% Amp B, and the cell suspension was incubated in a T‐25 flask at 37°C with 5% CO₂. Patient‐derived data were collected and utilized in accordance with the Research Ethics Committee Approval, decision number 2019.257.IRB2.079.

2.2. Immunofluorescence Staining of ULA and GelMA‐Pm Spheroids

2.2.1. Spheroids Collected From ULA Plates

Cells were collected from ULA plates and centrifuged at 200 × g. Supernatant was discarded and the spheroid containing pellet was gently washed with PBS once, and centrifuged again. Pellet was then resuspended in 4% PFA solution and kept in 4°C for 20 min. PFA solution was removed after another centrifuge. For blocking and as a staining solution, CellO‐IF (Cellorama) product was used. Cells were resuspended in 3–5 drops of CellO‐IF solution and incubated at room temperature for 45 min. Later, this same solution was enriched with 1:50 α‐Tubulin Rabbit Monoclonal Antibody (Cell Signaling) and incubated overnight. The next day, after addition of 3 mL PBS, the cells were pelleted and Anti‐rabbit Alexa Fluor 488 was diluted 1:50 in CellO‐IF solution and added to the cells. After 2 h of room temperature incubation, 3 mL of PBS was added to the staining solution and cells were pelleted again. CD133‐PE (Miltenyi Biotec, equal mixing of 130‐110‐962 and 130‐110‐962, 1:50–1:50 = 2:100) antibody was diluted 1:50 in CellO‐IF and incubation was done at 4°C for 20 min. Same amount of PBS addition and cell pelleting procedure was applied. Lastly, cells were resuspended in 30 µL of DAPI mounting media (VectaShield) and taken into a 35‐mm glass bottom confocal dish. Imaging was performed with a Leica confocal microscope. Objectives with magnification of 10× and 63× were used. The 63× objective was more effective in visualizing the CD133‐PE channel. DAPI (assigned blue color) and Alexa Fluor 488 (assigned green color) were visualized with PMT (Gain 800) detectors, and PE (assigned red color) were visualized with Hybrid (Gain 100) detectors. Ranges of the detectors were kept between required intervals. DAPI (405) light source emitted light between 1% and 10%, argon (488) light source emitted light between 5% and 20%, and PE (594) light source emitted light between 20% and 40%, generally. LUT mode showed no over‐excitation of any dye, and no stain was over or under exaggerated.

2.2.2. GelMA‐Pm Polymer Structure Immunofluorescence Application

Confocal dish was taken from the incubator, and media were removed. After a PBS wash, 1 mL 4% PFA was applied onto the polymer structures and kept in 4°C for 1 h. Later, the PFA was removed, and permeabilization solution, 0.1% TritonX‐100 in 1X DPBS, was added for 30 min in room temperature. Structures were washed once with PBS and then treated with the SuperBlock solution for 40 min in room temperature. After a PBS wash, structures were incubated with 1:50 α‐Tubulin Rabbit Monoclonal Antibody (Cell Signaling) in a PBS solution overnight at 4°C. Secondary antibody Alexa Fluor 488 was incubated for 2 h the next day after a PBS rinse. Secondary antibody incubation was terminated, and another PBS rinse was applied. CD133‐PE (Miltenyi Biotec, equal mixing of 130‐110‐962 and 130‐110‐962, 1:50‐1:50 = 2:100) conjugate antibody was kept on the structure for 30 min in 4°C and washed with PBS afterwards. Finally, DAPI mounting media was added to the polymers (100 µL) and taken to the confocal microscopy for imaging. Objectives with magnifications of 10× and 20× were used. The 20× objective was more effective in visualizing the CD133‐PE channel. DAPI (assigned blue color) and Alexa Fluor 488 (assigned green color) were visualized with PMT (Gain 800) detectors and PE (assigned red color9) was visualized with Hybrid (Gain 100) detectors. Ranges of the detectors were kept between required intervals. DAPI (405) light source emitted light between 1% and 10%, argon (488) light source emitted light between 5% and 20%, and PE (594) light source emitted light between 20% and 40%, generally. LUT mode showed no over‐excitation of any dye, and no stain was over or under exaggerated.

2.3. Scanning Electron Microscopy (SEM)

For structural characterization, acellular and cell‐laden GelMA hydrogels were collected at Day 0 and Day 14 of culture and processed for SEM imaging. Samples were fixed, dehydrated, and imaged at multiple magnifications to evaluate hydrogel surface morphology, culture‐associated microstructural remodeling, and cell–matrix interactions during long‐term 3D culture.

2.4. NCI‐SRB Viability Assay

Cells were plated in 96‐well plates (3000 cell/well) monolayer group directly, and spheroid group after reducing to single cells. They were grown 48 h before being treated with increasing concentrations of the compounds. After 72 h of incubation (5 days for Olaparib plates), cells were fixed with 10% (v/v) trichloroacetic acid for 1 h at 4°C. Then plates were washed five times with deionized water and left to dry. Next day, cells were stained with 0.4% (m/v) of sulforhodamine B in 1% acetic acid solution in the dark at room temperature for 30 min. The plates washed five times with 1% acetic acid before air‐drying. Bound SRB was dissolved in 10 mM Tris‐base solution. The absorbance was read in a plate reader at 515 nm. DMSO was used as a negative control. The experiment was performed in triplicates and biological replica plates were performed in triplicates as well.

2.5. Western Blot

Cells were collected from cultures to yield at least 5 × 10⁵ cells per group. For GelMA‐Pm cells, cells were first treated with Collagenase Type II for 15 min in the 37°C incubator with 5% CO₂ and centrifuged in 300 × g for 5 min to remove polymer structures and collect cells as a pellet. Pellets were collected and lysed with 1X RIPA Lysis Buffer, 10% proteinase inhibitor and 1% phosphatase inhibitor. Lysates were incubated on ice for 45 min, vortexing every 15 min. Samples were centrifuged at 13000 × g for 15 min. The protein concentration of the lysates was determined with the BCA assay (Thermo Scientific, 23225 Kit Protocol was followed). Cell lysates containing 20 µg protein were mixed with 4X Laemmli Sample Buffer and 1/6 β‐mercaptoethanol and then denatured for 7 min in 95°C. Samples were loaded to precast gels of 12 wells (4%–15% gradient) (BIORAD). After electrophoresis, the proteins were transferred to PVDF membranes with the Semidry Transfer System, followed by overnight incubation in a blocking solution (5% Blotting Grade Blocker in 1X TBS‐T (0.2% tween). Primary antibodies were used in a ratio of 1:500 to 1:1000 in 5% BSA‐TBS‐T, incubated for 1 h in room temperature (Cell Signaling). Secondary antibodies was used in 1:2000 dilution (Abcam). Different housekeeping proteins, including GAPDH, vinculin, calnexin, and α‐tubulin, were selected depending on target protein molecular weight, membrane region availability, transfer quality, and compatibility with simultaneous target detection on the same membrane. To minimize potential signal distortion associated with repeated stripping procedures, membranes were preferentially processed without extensive stripping whenever feasible. Loading controls were therefore chosen according to the most technically reliable and non‐overlapping membrane regions for each experiment. For visualization, chemiluminescence was performed with Immobilon Forte (Milipore), using LI‐COR Fc. Blots were cropped to highlight relevant bands. All uncropped full‐length membranes are provided in Figure S5. Quantification was done with ImageJ software.

2.6. Flow Cytometry for Cell Surface Markers

Cells were collected, washed with ice‐cold 1%FBS/PBS/2 mM EDTA solution and fixed with 4% PFA solution with gentle pipetting. Then cells were centrifuged and resuspended in their respective antibody mix (2 µL antibody + 98 µL 1% FBS/PBS Solution) and incubated at 4°C for 10 min. Samples were washed with DPBS three times and re‐suspended in 1% FBS/PBS/2 mM EDTA solution. Flow Cytometry (CytoFLEX) was used for analysis. All antibodies were conjugate antibodies (Miltenyi Biotec) requiring no secondary staining.

2.7. RT‐qPCR Reaction

Primer mixes were prepared as 90 µL NF water + 5 µL Forward Primer (100 µM) + 5 µL Reverse Primer (100 µM). Prepared mixture for each primer; 10µL SYBR, 7 µL NF water, 1 µL primer mixes, 2 µL cDNA was loaded to the 96‐well qPCR plates. Plates were spun down before loading onto the Light Cycler 480 (Roche) with 65°C melting curve and 50 cycles (Table 1).

TABLE 1.

RT‐qPCR primer sequences.

Gene	Forward primer	Reverse primer
β‐Actin	CACCATTGGCAATGAGCGGTTC	AGGTCTTTGCGGATGTCCACGT
SOX‐2	AACCAGCGCATGGACAGTTA	GACTTGACCACCGAACCCAT
ALDH1A1	CTGTGTTCCAGGAGCCGAAT	AGCATCCATAGTACGCCACG
NANOG	TGATTTGTGGGCCTGAAGAAA	TGGTGGTAGGAAGAGTAAAG
MYC	CGTCCTCGGATTCTCTGCTC	GCTGGTGCATTTTCGGTTGT

2.8. RNA Isolation

2.8.1. Cell Lines

For cell lines and primary patient‐derived cells grown as 2D, cells were trypsinized with 0.25% trypsin‐EDTA. For 3D‐ULA cells, only collection and pelleting was applied without trypsin. Pellets obtained with 300 × g centrifugation and pellets were stored at −80°C until isolation. For isolation, Total RNA Extraction kit (Macherey Nagel) protocol was followed, and at the end of the experiment, Nanodrop reads for concentrations and purities were applied.

2.8.2. GelMa‐Pm

GelMA‐Pm encapsulates were treated with Collagenase Type II (Gibco, 1 mg/mL) for 15 min in the 37°C incubator with 5% CO₂. After polymer structures were dissipated and cells were exposed, cells were centrifuged to obtain a pellet. Pellet was resuspended in TRIzoL (Qiagen) and kept at −20°C overnight. Frozen cell pellets were thawed on ice the next day, and 200 µL chloroform was added and vortexed until homogenous. Mixture was sat in room temperature for 5 min until phase separation occurred. It was then centrifuged at 12000 × g for 15 min at 4°C, without the upper phase touching the middle phase. After this step, Total RNA Extraction kit (Macherey Nagel) protocol was followed, and at the end of the experiment, Nanodrop reads for concentrations and purities were applied.

2.8.3. Tissue

Tissue samples were removed from the cryovial containing the RNAlater solution (Invitrogen) and transferred to new 2 mL Eppendorf tubes. Then, 300 µL of 1X DNA/RNA Shield (Zymo Research) was added to each tissue sample, and the tissues were minced using Rotor‐Stator Homogenizer until no fragment was larger than half the diameter of the homogenization probe. Homogenization was carried out in 15–20 s intervals with 5‐s pauses in between, for a total duration of 60 s. During these intervals, the polytron speed was reduced, and the probe was gently tapped against the tube wall. After tissues were completely homogenized, samples were centrifuged at 12.000 × g for 10 min at +4°C, and the supernatant was collected. Following homogenization, Total RNA of tissues was isolated according to the Zymo Quick‐RNA Miniprep Plus kit protocol, and at the end of the experiment, Nanodrop reads for concentrations and purities were applied.

2.9. Statistical Analysis and Utilized Softwares

Statistical analyses were calculated with GraphPad Prism 10.4.1 using tests ANOVA, t‐test or Correlation Matrix. p value symbols throughout the article are (*) for 0.1 > p value, (**) for 0.05 > p value, (***) 0.01 > p value. Therefore, readers should acknowledge that wherever 1 star significance (*) is indicated, confidence interval is 90%.

All graphics contain error bars to respective standard deviations, and graphs show mean values of technical and biological replicates. Graphics were drawn in GraphPad and BioRender, western blot membranes were imaged and exported through ImageStudio and quantified with ImageJ, confocal microscopy images were imaged, snapshotted, cropped, and modified with LAS X software. RT‐qPCR primers were designed with Blast‐Primer tool.

3. Results

3.1. Optimization and Spheroid Maturation in ULA Cultures of HGSOC Cell Lines

We first characterized spheroid formation and maturation dynamics in scaffold‐free ULA cultures using the HGSOC cell lines OVCAR‐3 and OVSAHO. In both lines, cells initially formed loose aggregates by Day 7, which progressively compacted into spheroids by Day 14 (Figure 1a,b). This timeframe is consistent with previous reports describing spheroid formation across multiple tumor types, including pancreatic, breast, and prostate cancers, typically achieved within 1–2 weeks using hanging drop or U‐bottom systems (Figure 1a,b). While such small‐volume approaches are suitable for short‐term viability assays, they are inherently limited for applications requiring higher cell yields. In contrast, 6‐well ULA plates enabled the generation of sufficient spheroid material for downstream CSC‐focused analyses, including protein and RNA profiling and higher‐throughput applications such as CRISPR‐based screens.

FIGURE 1.

Spheroid formation, maturation, and functional CSC enrichment in HGSOC cell lines cultured under ultra‐low attachment (ULA) conditions. Representative brightfield microscopy images of OVCAR‐3 and OVSAHO cell lines cultured in ULA plates at relevant time points (4× objective). Yellow arrows indicate highly compartmentalized OVCAR‐3 spheroids, whereas red arrows denote overcrowded aggregates observed at extended culture durations (a). Schematic illustration summarizing spheroid morphologies observed under ULA conditions, including loose aggregates, compact spheroids, and compartmentalized spheroids (b). Quantitative analysis of spheroid diameter distribution and representative size comparison of highly compartmentalized OVCAR‐3 spheroids illustrating intra‐well heterogeneity within ULA cultures at Day 48 (10× objective) (c). Representative confocal microscopy images of an OVCAR‐3 spheroid at Day 48 including brightfield image, confocal z‐sections acquired at ∼2 µm intervals, and corresponding z‐stack projection (63× objective). Blue = DAPI (nuclei), green = α‐tubulin. Individual channels are shown separately to facilitate visualization of marker distribution. CD133‐PE monolayer negative controls and uncropped confocal images are provided in Figure S1 (d). Representative confocal microscopy images of additional aggregates and spheroids from the same culture (63X objective). Blue = DAPI, green = α‐tubulin, red = CD133‐PE. Individual channels are shown separately. CD133‐PE monolayer negative controls and uncropped images are provided in Figure S1 (e). Growth inhibition curves and corresponding IC50 bar graphs for carboplatin, niraparib, paclitaxel, olaparib, cisplatin, and doxorubicin comparing OVCAR‐3 monolayer cultures (black bars) and ULA spheroids (gray bars). Viability was assessed using the NCI‐SRB assay (f). Data are presented as mean ± SD from three independent biological replicates. For SRB assays, each biological replicate contained technical triplicates. Statistical significance was determined using unpaired two‐tailed Student's t‐tests or one‐way ANOVA where appropriate. *p < 0.05, **p < 0.01, ***p < 0.001.

By Day 21, OVCAR‐3 spheroids exhibited further morphological maturation, developing into distinct “compartmentalized spheroids” characterized by visible proliferative, quiescent, and necrotic zones. In contrast, OVSAHO cultures remained predominantly as loose aggregates without clear structural compartmentalization. Compact spheroids displayed uniform density, whereas compartmentalized spheroids retained discernible internal architecture. This classification into loose, compact, and compartmentalized morphologies aligns with established descriptions of multicellular tumor spheroids in the literature [50, 51, 52, 53], and is supported by prior mass spectrometry analyses reported by Acland et al. [54].

By Day 28, OVCAR‐3‐compartmentalized spheroids reached approximately 200 µm in diameter; however, cultures remained heterogeneous, with smaller spheroids ranging from 50 to 100 µm coexisting within the same well (Figure 1c). Prolonged culture beyond this period (> 48 days) resulted in overcrowding and the emergence of opaque, necrotic aggregates (Figure 1a, red arrow), indicating declining culture integrity. Based on these observations, a culture duration of approximately 28 days was defined as the optimal endpoint for ULA spheroids, balancing morphological organization, cell viability, and CSC enrichment potential.

Confocal imaging confirmed both structural integrity and CSC marker enrichment in ULA spheroids. Representative z‐stack projections (Figure 1d,e) demonstrate CD133‐PE and α‐tubulin‐Alexa488 staining, with DAPI counterstaining. CD133‐PE images of monolayer cultures are included as a negative control, along with unmerged versions of the images in Figure S1, confirming that CD133 expression is specifically enriched under spheroid conditions. Notably, despite the absence of compartmentalized morphology in OVSAHO cultures, both OVCAR‐3 and OVSAHO ULA spheroids showed significant upregulation of CSC‐associated intracellular proteins, including SOX2, OCT4, NANOG, c‐MYC, and ALDH1A1, after approximately 28 days of culture (Figure 3a). These findings indicate that CSC enrichment can occur independently of overt structural compartmentalization, although the latter may better approximate in vivo tumor heterogeneity.

FIGURE 3.

Comparative benchmarking of CSC and EMT‐associated traits in HGSOC cells cultured using ULA and photopatterned GelMA platforms. Representative western blot images and corresponding quantitative bar graphs showing expression of CSC‐associated proteins SOX‐2, ALDH1A1, NANOG, and OCT‐4 in OVCAR‐3 and OVSAHO cultures. Protein levels were normalized to Vinculin and Calnexin, with representative loading control bands shown. Different loading controls were selected depending on target abundance, membrane compatibility, and subcellular localization. Quantification was performed using ImageJ (a). Representative flow cytometry pseudo‐dot plots with gates defined using unstained controls, and bar graphs summarizing the percentage of CD133‐PE, CD44‐FITC, CD117‐PE, and CXCR4‐FITC positive populations. Representative gating strategy is provided in Figure S4 (b). Representative western blot images and quantitative bar graphs of EMT‐associated proteins Slug, Snail, TWIST1, vimentin, and fibronectin in OVCAR‐3 cultures (normalized to calnexin). E‐cadherin and N‐cadherin were additionally analyzed (normalized to α‐tubulin), with quantitative N‐cadherin/E‐cadherin ratios shown (c). Representative western blot analysis of stemness‐associated proteins in OVCAR‐3 cells cultured as 2D monolayer, 2D monolayer in spheroid enrichment medium, UV‐treated monolayer, ULA spheroids, and photopatterned GelMA (GelMA‐Pm) spheroids. Due to limited cell yield obtained from GelMA‐Pm cultures, only SOX‐2 and ALDH1A1 expression could be evaluated in these samples. Protein expression was normalized to GAPDH. Corresponding densitometric quantification is shown on the right. Full‐length uncropped blots are provided in Figure S5. (d). RT‐qPCR analysis of the stemness‐associated genes SOX‐2, NANOG, ALDH1A1, and c‐MYC in OVCAR‐3 cells cultured under 2D monolayer, 2D + sphere medium, 2D + UV, photopatterned GelMA (GelMA‐Pm), and ULA spheroid conditions. Gene expression levels were normalized to β‐actin and are presented relative to 2D monolayer controls (set to 1). Values above 1 indicate increased expression relative to monolayer cultures. Data represent mean ± SD from three independent biological replicates (n = 3). Statistical significance was determined using multiple t‐tests; significance symbols are defined as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***) (e). Data are presented as mean ± SD from three independent biological replicates. Statistical significance was determined using unpaired two‐tailed Student's t‐tests or one‐way ANOVA where appropriate. * p < 0.05, ** p < 0.01, *** p < 0.001. Western blot images were cropped for clarity; non‐adjacent lanes from the same membrane are indicated by spaces. Full‐length blots are provided in Figure S5.

To assess whether prolonged ULA culture translated into functional consequences, we compared drug responses between monolayer and ULA spheroid cultures. Growth inhibition curves and IC50 analyses (Figure 1f) revealed that OVCAR‐3 ULA spheroids acquired significant resistance to multiple clinically relevant agents, including carboplatin, niraparib, paclitaxel, olaparib, cisplatin, and doxorubicin. Together, these data demonstrate that ULA culture supports progressive spheroid maturation and functional CSC enrichment, establishing approximately 28 days as an optimal culture duration for downstream CSC‐focused studies.

3.2. Optimization of Photopatterned GelMA Hydrogels for Rapid Spheroid Formation in HGSOC Models

We next optimized spheroid formation using photopatterned GelMA hydrogels. Initial experiments employed large GelMA–Irgacure droplets (∼100 µL, 2 × 10⁶ cells/mL) crosslinked under UV exposure for 30 s at a distance of 5 cm. Under these conditions, spheroid formation was inefficient, with only small structures emerging after approximately 1 month and overall cell viability remaining low. In addition, culture medium retained a red coloration for extended periods, suggesting limited metabolic activity and nutrient utilization by encapsulated cells.

To improve nutrient diffusion and spatial confinement, photomasks were introduced between the UV source and GelMA droplets, generating smaller and more uniform encapsulation compartments (Figure 2b). This modification enhanced cell viability and promoted spheroid initiation, although proliferation remained limited under FBS‐depleted spheroid enrichment conditions. We therefore systematically optimized both initial cell density and UV exposure duration. Optimal conditions were identified as 4 × 10⁶ cells/mL for OVSAHO and 3 × 10⁶ cells/mL for OVCAR‐3, combined with 50 s of UV exposure at 5 cm (6.25 W/cm²). Photomasks containing square transparent features (∼0.64 mm²) produced reproducible hydrogel structures that were readily visualized on glass‐bottom culture dishes (Figure 2b).

FIGURE 2.

Optimization of photopatterned GelMA hydrogels for spheroid formation and CSC enrichment in HGSOC cell lines. Representative brightfield microscopy images of photopatterned GelMA (GelMA‐Pm) cultures of OVCAR‐3 and OVSAHO cells (upper panels: 4× objective; lower panels: 10× objective), showing spheroid formation under optimized conditions (a). Schematic overview of the GelMA‐Pm fabrication workflow and representative image of square photomask‐generated polymerized GelMA structures visible to the naked eye (b). Quantitative comparison of spheroid number and spheroid diameter per mm² between ULA and GelMA‐Pm cultures (c). Representative brightfield microscopy images illustrating approximate spheroid diameter distributions within photopatterned GelMA cultures at Day 14, including representative OVCAR‐3 spheroids (∼100 µm) and smaller OVSAHO spheroids (∼50 µm) (d). Representative confocal microscopy images of GelMA hydrogels containing spheroids from OVCAR‐3 and OVSAHO cultures, demonstrating cell viability and spatial distribution (10× objective, with selected regions magnified to 20X as indicated by yellow boxes). Blue = DAPI (nuclei), green = α‐tubulin, red = CD133‐PE. Corresponding brightfield and fluorescence images of identical fields are shown (e). Data are presented as mean ± SD from three independent biological replicates. Statistical significance was determined using unpaired two‐tailed Student's t‐tests or one‐way ANOVA where appropriate. * p < 0.05, ** p < 0.01, *** p < 0.001.

Under these optimized conditions, spheroids became detectable as early as Day 9, and compact spheroids were consistently observed by Day 14 (Figure 2a, yellow arrows). Quantitative analysis demonstrated that photopatterned GelMA cultures produced a higher number of spheroids per area compared with ULA cultures (Figure 2c).

Representative size‐reference images of GelMA‐derived spheroids at Day 14 further demonstrated reproducible differences between the two HGSOC models (Figure 2d). OVCAR‐3 cultures predominantly formed moderately sized spheroids with diameters frequently around ∼100 µm, whereas OVSAHO cultures tended to generate smaller but more numerous spheroids, commonly around ∼50 µm in diameter. In contrast, prolonged ULA cultures of OVCAR‐3 exhibited broader spheroid size variability, including both small aggregates and larger compartmentalized spheroids. These observations support distinct growth dynamics between scaffold‐free and ECM‐based 3D culture conditions

Confocal imaging at Day 14 confirmed the presence of viable spheroids, with DAPI nuclear staining and clear CD133‐PE expression (Figure 2e). α‐Tubulin‐Alexa488 staining provided structural reference, although signal intensity varied with imaging depth. Both 10× and 20× objective views are shown, with higher magnification enabling improved visualization of CD133‐PE. CD133‐PE signal appeared more prominent in OVCAR‐3 spheroids relative to OVSAHO, a difference explored further in subsequent stemness analyses.

Importantly, photopatterned GelMA enabled OVSAHO—otherwise limited to loose aggregates in ULA culture—to form compact spheroids, suggesting that ECM‐derived cues and the structural properties of the hydrogel contribute to enhanced cellular clustering. OVCAR‐3 spheroids also formed readily under these conditions and were generally larger than those derived from OVSAHO. However, extended GelMA culture beyond 14 days was associated with progressive matrix loosening, microstructural remodeling, and frequent cell escape from the hydrogel boundaries, ultimately leading to mixed 2D–3D cultures (Figure S2). Based on these observations, Day 14 was defined as the optimal endpoint for photopatterned GelMA cultures, at which point spheroids exhibited high viability, reproducible morphology, and measurable CSC marker expression suitable for downstream analyses.

3.3. Benchmarking CSC and EMT‐Associated Traits Across ULA and Photopatterned GelMA Platforms

To directly compare the biological outputs of the two 3D culture platforms, we next benchmarked CSC‐associated surface markers, intracellular stemness regulators, and epithelial–mesenchymal transition (EMT) signatures. CSC traits were defined by the combined assessment of surface marker expression, intracellular stemness proteins, and functional spheroid behavior.

As shown in Figure 3a, ULA‐cultured spheroids from both OVCAR‐3 and OVSAHO displayed increased expression of ALDH1A1, SOX2, NANOG, and OCT4 relative to monolayer controls. In OVSAHO, the increase in ALDH1A1 was more modest but reached statistical significance at 90% confidence. Flow cytometric analysis of CSC surface markers (CD133, CD44, CD117, and CXCR4) revealed pronounced enrichment in OVCAR‐3 spheroids for all markers tested (Figure 3b). In contrast, OVSAHO showed a significant increase primarily in CD117 expression. Notably, CD133 positivity in OVCAR‐3 spheroid populations frequently reached 70%–80% across multiple replicates, with remaining experiments showing enrichment to 50%–60%. This variability falls within the expected biological range for CSC enrichment under 3D culture conditions, particularly in the presence of growth factor and insulin supplementation, and is consistent with prior reports of ∼74% CD133 positivity in OVCAR‐3 spheroids compared with ∼6% in adherent cultures [46, 55]. To control for potential effects of culture medium and UV exposure, we included monolayer cultures maintained in spheroid enrichment medium and monolayers exposed to the same UV conditions used for GelMA photopatterning. These controls showed only modest changes in selected markers, confirming that robust CSC enrichment required 3D culture. Figure S3 further demonstrates that while soluble factors can influence certain markers, consistent CSC enrichment in OVCAR‐3 necessitates ULA‐based spheroid formation. For all flow cytometry results gating strategies, singlet selection and unstained samples are shown in Figure S4.

EMT marker analysis revealed significant increases in Slug, Snail, TWIST1, Vimentin, and Fibronectin in spheroid cultures (Figure 3c). E‐cadherin levels remained largely unchanged, whereas N‐cadherin showed a significant increase in OVCAR‐3 ULA spheroids. Quantitative analysis of the N‐cadherin/E‐cadherin ratio confirmed a shift toward a more mesenchymal phenotype (Figure 3c). Blots were cropped for clarity, with non‐adjacent lanes from the same membrane indicated; full‐length membranes are provided in Figure S5.

Comparative analysis across platforms is shown in Figure 3d,e. Western blot and RT‐qPCR analyses revealed significant enrichment of ALDH1A1, SOX2, NANOG, and c‐MYC in both ULA and photopatterned GelMA cultures relative to controls. Because GelMA encapsulation yielded fewer cells per replicate, qPCR analyses were performed with increased replicate numbers to complement protein‐level data. UV exposure or spheroid enrichment medium alone had minimal effects, with the exception of SOX2 in OVCAR‐3. Differences between cell lines may reflect intrinsic genetic backgrounds, including BRCA status (OVCAR‐3 BRCA wild‐type versus homozygous BRCA deletion in OVSAHO). Collectively, these results demonstrate that both ULA and photopatterned GelMA platforms robustly enrich CSC and EMT‐associated traits, with ULA favoring maintenance of intrinsically stem‐like populations and GelMA efficiently inducing stem‐like features across models.

3.4. Validation of ULA and Photopatterned GelMA Platforms Using Patient‐Derived HGSOC Samples

To evaluate the translational relevance of the two 3D culture platforms, we extended our analysis to patient‐derived HGSOC samples. Four patient cultures were established (two from tumor tissue and two from ascites), and brightfield images of ULA and photopatterned GelMA spheroids are shown in Figure 4a. Primary cells were initially expanded under 2D conditions to obtain sufficient material for downstream assays, reflecting the limited yield and slow proliferation typical of patient‐derived ovarian cancer cells.

FIGURE 4.

Validation of ULA and photopatterned GelMA platforms using patient‐derived HGSOC primary cells. Representative brightfield microscopy images of patient‐derived primary HGSOC cells cultured under ULA or photopatterned GelMA conditions (upper panels: 4X objective; lower panels: 10× objective) (a). IC50 values for carboplatin, niraparib, paclitaxel, and olaparib in 2D monolayer cultures of four patient samples, determined using the NCI‐SRB assay (b). Baseline SOX‐2 expression in patient tumor tissues measured by RT‐qPCR and normalized to β‐actin (c). RT‐qPCR analysis of SOX‐2, NANOG, ALDH1A1, and c‐MYC expression in patient‐derived cells cultured as 2D monolayers, ULA spheroids, or photopatterned GelMA spheroids, normalized to β‐actin (d). Correlation matrix heatmap depicting relationships between drug response profiles, spheroid numbers, and CSC marker expression, with correlation coefficients ranging from −1.0 (negative correlation) to +1.0 (positive correlation) (e). Data are presented as mean ± SD from three independent biological replicates unless otherwise indicated. Statistical significance was determined using unpaired two‐tailed Student's t‐tests, one‐way ANOVA, or Pearson correlation analysis where appropriate. *p < 0.05, **p < 0.01, ***p < 0.001.

Drug sensitivity profiling was performed at the 2D stage using an SRB assay with Carboplatin, Paclitaxel, Olaparib, and Niraparib (Figure 4b). To relate drug response to intrinsic stemness, SOX2 expression was measured from RNA isolated directly from patient tissues prior to culture. Partial correlations were observed between SOX2 expression and drug resistance. Patient 1, exhibiting the lowest SOX2 expression, was the most sensitive to three of four agents, whereas Patient 2—clinically deemed inoperable and succumbing rapidly to disease—displayed the highest SOX2 levels and marked resistance, particularly to Olaparib and Paclitaxel. Patient 3 showed pronounced resistance to Carboplatin, while Patient 4 exhibited less consistent correlations. BRCA mutation status was unavailable for all patients.

CSC enrichment under 3D conditions was assessed by qPCR analysis of SOX2 and NANOG expression (Figure 4d). In most patients, photopatterned GelMA induced stronger CSC marker enrichment than ULA, with the exception of Patient 2, in whom NANOG expression decreased despite high baseline SOX2. Correlation analysis (Figure 4e) revealed an inverse association between spheroid number and both drug resistance and SOX2 expression, indicating that spheroid abundance alone does not necessarily reflect CSC strength or chemoresistance in primary cultures.

Overall, both ULA and photopatterned GelMA platforms effectively enriched CSC traits in patient‐derived HGSOC samples. ULA cultures largely preserved CSC features in proportion to baseline stemness, whereas photopatterned GelMA hydrogels enhanced CSC marker expression even in samples with initially low stemness. No significant differences were observed between tumor‐derived and ascites‐derived cells in their capacity to form spheroids or respond to 3D culture, underscoring the complementary utility of both platforms for translational HGSOC modeling.

4. Discussion

Optimizing reliable three‐dimensional (3D) culture platforms for CSC enrichment in HGSOC is essential for understanding mechanisms underlying tumor progression, therapy resistance, and disease recurrence [56]. In this study, we benchmarked two widely used 3D approaches—ULA culture and photopatterned gelatin methacryloyl (GelMA‐Pm) hydrogels—and demonstrate that both systems enrich CSC‐associated traits, although with distinct biological outputs and practical advantages. Our findings highlight that CSC‐associated phenotypes in HGSOC are strongly influenced by both intrinsic tumor cell properties and culture platform architecture, emphasizing that model selection should be guided by the intended biological question and translational application.

Although ULA is an established scaffold‐free method, our results show that it remains a highly effective CSC enrichment platform when sufficient culture duration is allowed. In OVCAR‐3 cells, nearly 4 weeks of ULA culture were required to generate large, highly compartmentalized spheroids accompanied by pronounced CSC marker enrichment, whereas OVSAHO cultures predominantly formed loose aggregates even after extended culture. These observations underscore that spheroid morphology is cell‐line dependent and should not be overgeneralized across HGSOC models. Functional assays nonetheless confirmed CSC‐associated enrichment in both OVCAR‐3 and OVSAHO, demonstrating that visible compartmentalization is not strictly required for CSC traits to emerge, consistent with prior OCSC models [57]. At the same time, in our dataset, compartmentalized OVCAR‐3 spheroids coincided with the strongest CSC surface marker expression and multidrug resistance, suggesting that spheroid architecture may reinforce therapy‐resistant phenotypes in specific contexts. Similar layered or compartmentalized spheroid structures have been reported in breast, colorectal, and pancreatic cancer models cultured in 3D systems [58, 59]. Whether such architectural features directly drive CSC phenotypes or reflect downstream consequences of prolonged 3D growth remains an open question. These findings also suggest that short‐term spheroid assays may underestimate CSC potential, particularly in models such as OVCAR‐3 that have previously been described as poor or loose spheroid formers under abbreviated culture conditions [57, 60]. In contrast, GelMA‐Pm enabled compact spheroid formation within 2 weeks and was particularly effective in OVSAHO, a model that otherwise showed limited scaffold‐free spheroid formation. This suggests that extracellular matrix (ECM)‐associated cues and hydrogel‐supported spatial confinement can promote spheroid organization in models with weak intrinsic aggregation capacity. Importantly, SEM and brightfield‐based characterization further supported time‐dependent matrix remodeling and culture‐associated loosening of the GelMA system, consistent with our observation that extended culture beyond 14 days resulted in cell escape and mixed 2D–3D growth patterns. Therefore, while ULA is advantageous for long‐term spheroid maturation, GelMA‐Pm provides a shorter, ECM‐assisted culture window for reproducible compact spheroid formation and CSC‐associated marker induction.

While both platforms promoted CSC‐associated features, the magnitude and qualitative profile of enrichment differed. OVCAR‐3 spheroids generated under ULA conditions consistently displayed increased expression of SOX2, NANOG, ALDH1A1, and CD133, with CD133 positivity frequently reaching 70%–80%. Although this proportion appears high relative to the small stem‐like fractions typically reported in adherent cultures, it aligns with previous studies showing that spheroid culture can expand the CD133⁺ population in OVCAR‐3 cells from ∼6% to ∼74% [55]. This enrichment was further supported by coordinated induction of epithelial–mesenchymal transition (EMT)‐associated markers. Upregulation of Slug, Snail, TWIST1, Vimentin, and Fibronectin, together with increased N‐cadherin and a higher N/E‐cadherin ratio, is indicative of partial EMT and reinforces the established link between 3D growth, EMT activation, and CSC‐associated phenotypes [61].

A key interpretive consideration is whether increased stemness marker expression reflects selective enrichment of pre‐existing CSC‐like populations or induction of CSC‐like states in non‐CSC populations. Our findings suggest that both mechanisms may contribute, but their relative importance likely differs by platform. ULA culture may preferentially select for or sustain intrinsically stem‐like subpopulations that tolerate anchorage‐independent, low‐serum, growth factor–supplemented conditions. In contrast, GelMA‐Pm may more strongly promote phenotypic plasticity through ECM‐associated cues, spatial confinement, and matrix remodeling, particularly in lower‐stemness or weakly aggregating models such as OVSAHO. This distinction is important because induction‐based systems may not always serve as direct predictors of baseline clinical drug response. Instead, GelMA‐Pm may be especially useful as a mechanistic tool to study ECM‐driven CSC plasticity, while ULA may be more suitable for scalable enrichment and maintenance of intrinsic CSC‐associated states.

Differential responses between OVCAR‐3 and OVSAHO may also reflect their distinct genetic backgrounds, as OVCAR‐3 is BRCA1/2 wild‐type whereas OVSAHO harbors a homozygous BRCA2 deletion [28, 62]. Notably, UV exposure inherent to GelMA‐Pm fabrication enhanced CSC marker expression more strongly in OVSAHO, potentially reflecting impaired DNA damage repair capacity in this line. While this observation requires further validation, it highlights how molecular context can modulate responses to engineered 3D platforms and underscores the importance of aligning model choice with genetic features when studying CSC biology [63].

Extending these platforms to patient‐derived primary HGSOC cells further underscored both their translational relevance and practical limitations. Primary ovarian cancer cells proliferate slowly and are typically available in limited quantities, necessitating initial 2D expansion prior to 3D culture. This constraint restricted large‐scale drug testing directly in 3D systems and prompted the use of 2D drug sensitivity assays as a baseline comparator. Despite this limitation, intrinsic SOX2 expression in patient tissues partially correlated with chemoresistance, exemplified by Patient 2, who exhibited high SOX2 levels, marked drug resistance, and poor clinical outcome. Notably, spheroid number alone did not correlate with CSC marker expression or drug response, indicating that spheroid abundance is not a reliable surrogate for CSC phenotype [64, 65]. In patient‐derived cultures, GelMA‐Pm generally enhanced CSC marker expression even in samples with lower baseline stemness, whereas ULA cultures more closely preserved intrinsic CSC‐associated patterns. These differences further support the interpretation that ULA may better model baseline CSC representation, while GelMA‐Pm may serve as an ECM‐assisted CSC induction platform for mechanistic studies.

Our findings also place ULA and GelMA‐Pm within the broader landscape of ovarian cancer 3D models, including hanging drop systems and organoid cultures embedded in basement membrane extracts such as Matrigel. Hanging drop methods are useful for rapid spheroid generation and short‐term drug response studies due to their simplicity and ability to produce relatively uniform spheroids. However, they are generally limited by low throughput, restricted media exchange, and difficulties in generating sufficient biomass for downstream applications such as transcriptomic, proteomic, or CRISPR‐based analyses. Similarly, Matrigel‐based organoid systems preserve important aspects of tumor heterogeneity and tissue architecture and are increasingly used in personalized drug screening approaches [66, 67, 68]. Nevertheless, these systems remain biologically undefined, compositionally variable, costly, and less compatible with controlled biofabrication strategies or reproducible mechanical tuning. In contrast, ULA cultures provide a scalable and relatively inexpensive scaffold‐free platform for long‐term spheroid maturation and CSC enrichment, while photopatterned GelMA‐Pm hydrogels offer defined composition, spatial control, and ECM‐associated microenvironmental cues. Rather than functioning as competing technologies, these 3D systems should be considered complementary tools with distinct utilities: organoids preserve tissue‐level architecture and heterogeneity, hanging drop systems support rapid spheroid formation, ULA enables scalable CSC enrichment for downstream functional assays, and photopatterned GelMA facilitates controlled ECM mimicry and reproducible induction of stem‐like phenotypes.

This study has several strengths, including the direct comparison of two widely used 3D culture platforms under matched conditions, the integration of immortalized cell lines with patient‐derived samples, and the evaluation of CSC enrichment across multiple layers of analysis encompassing morphology, surface markers, intracellular stemness regulators, EMT signatures, and drug response. At the same time, several limitations should be acknowledged. First, the study was conducted exclusively using in vitro systems and did not include in vivo validation of tumorigenicity, metastatic behavior, stemness maintenance, or therapeutic response. Future studies incorporating xenograft or patient‐derived xenograft (PDX) models will therefore be important to determine how accurately these platforms recapitulate CSC‐driven tumor behavior in vivo. Second, although both ULA and GelMA‐Pm cultures partially mimic aspects of the tumor microenvironment, they do not fully reproduce the complex multicellular ecosystem of HGSOC, including stromal, immune, endothelial, and mesothelial interactions known to influence CSC plasticity and therapeutic resistance. Integration with co‐culture systems [69], organoid models [70], or immune/stromal components [71] may further improve physiological relevance.

Additional technical limitations inherent to spheroid‐based systems must also be considered, including size heterogeneity, nutrient gradients, and reproducibility challenges [60]. Limited cell numbers from primary cultures constrained extensive 3D drug testing, and GelMA‐Pm cultures required re‐encapsulation beyond 2 weeks due to hydrogel loosening and cell escape. Furthermore, BRCA mutation status was unavailable for each patient‐derived sample, limiting our ability to determine how homologous recombination deficiency or DNA repair capacity may have influenced spheroid behavior, CSC enrichment, or therapeutic response. Given the relationship between BRCA‐associated DNA repair defects, platinum sensitivity, PARP inhibitor response, and stemness‐associated phenotypes in HGSOC, future studies incorporating molecular stratification of patient samples will be important for improving the translational and predictive value of these models. Finally, while GelMA‐Pm efficiently enhanced CSC‐associated traits, particularly in lower‐stemness models, such induction‐based systems may partially exaggerate therapy‐resistant phenotypes relative to baseline patient tumors. Therefore, GelMA‐Pm should be interpreted primarily as a mechanistic platform for studying ECM‐associated CSC plasticity unless further validated in larger patient cohorts and in vivo settings.

Taken together, our results demonstrate that ULA and photopatterned GelMA are not interchangeable but rather complementary 3D platforms for modeling ovarian cancer stemness. ULA cultures are particularly advantageous for long‐term CSC maintenance, spheroid maturation, EMT‐associated phenotypes, and scalable downstream applications, whereas photopatterned GelMA platforms are especially effective for rapid ECM‐assisted spheroid formation and CSC induction in otherwise weakly aggregating HGSOC models. As engineered matrices continue to evolve to overcome the variability of natural ECMs [72], the strategic selection and combination of such platforms may improve the translational relevance of preclinical HGSOC models and support future development of CSC‐focused therapeutic strategies. However, further integration with multicellular tumor microenvironment models, molecularly stratified patient cohorts, and in vivo validation studies will be necessary to establish their predictive value for translational drug evaluation and precision oncology applications.

Author Contributions

D.D.Y., F.O., I.D.S., and E.S. contemplated the experimental processes. F.O. and M.K. conducted GelMA preparation, encapsulation, isolation, and analysis experiments. I.S.C. and G.Y.D. isolated primary patient‐derived cells from specimens collected surgically by B.G., D.V., and/or Ç.T. E.M.A. performed RNA isolation from patient tissues and archived them. D.D.Y conducted and analyzed all remaining experiments. I.D.S., F.O., and E.S. interpreted and recapitulated the results. All authors reviewed the manuscript.

Funding

No funding was received for this research.

Ethics Statement

The study was conducted in accordance with the relevant guidelines and regulations, and ethical approval for the study was granted from the Koç University Ethics Committee, Turkey (2019.257.IRB2.079). All patients provided written informed consent to participate in the study.

Consent

Written informed consent for publication of anonymized clinical data was obtained from all participants.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: biot70287‐sup‐0001‐SuppMat.zip.

Data Availability Statement

All data generated or analyzed during this study are included in this published article and its Supporting Information files. Raw data supporting the findings of this study are available from the corresponding author upon reasonable request.