An Advanced 3D Model of Vascularized Epithelial Ovarian Cancer in a Tumor-on-a-Chip System Based on Multi-Cell Culture

Highlights

What are the main findings?

• New EOC-on-a-chip model mimics heterogeneous and vascularized tumor tissue
• Complex model of ovarian cancer secretes proangiogenic factors (VEGF)

What is the implication of the main finding?

• HUVECs migrate toward an EOC-mimicking structure
• Long-term co-culture mimics inflammation and TME remodeling seen in tumor progression

Abstract

Epithelial ovarian cancer (EOC) is a highly lethal malignancy characterized by significant heterogeneity and poor prognosis due to late-stage diagnosis and chemotherapy resistance. Traditional two-dimensional (2D) models fail to replicate the complexity of the tumor microenvironment (TME), necessitating the development of advanced in vitro systems. Here, we present a novel microfluidic tumor-on-a-chip (ToC) system that accurately models key features of EOC, including heterogeneity and vascularization. The developed cellular model was evaluated for functionality. It was demonstrated that endothelial cells of blood vessels within a collagen matrix successfully migrated toward the cancerous tissue, while the multicellular and multilayered tumor construct secreted pro-angiogenic factors. Additionally, long-term culture conditions induced inflammatory responses, mimicking in vivo tumor progression. This innovative platform enables precise investigations into EOC biology, angiogenesis, and TME interactions. Furthermore, it holds significant potential for drug screening, assessing therapeutic efficacy, and advancing personalized oncology approaches.

1. Introduction

Cancers of the female reproductive system are the leading causes of cancer-related deaths in women. Ovarian cancer is the eighth most common cancer in women and is characterized by high mortality. According to WHO data, in 2022 alone, ovarian cancer caused over 207,000 deaths worldwide [1]. Most ovarian cancers are diagnosed at an advanced stage, which results in a five-year survival rate of only 15%. Epithelial ovarian cancer (EOC) is one of the most common types of ovarian cancer. It is characterized by the presence of cancerous tissue in the epithelial layer of the ovarian tissue. Cancer cells penetrate the basement membrane and interact with the stromal layer, creating a specific and heterogeneous tumor microenvironment (TME) [2,3]. Late diagnosis of EOC and, thus, a bad prognosis accompanied by low efficacy of chemotherapy are the main problems the healthcare world is facing. Therefore, new, advanced preclinical models and appropriate analytical procedures are desired to decipher the cellular and molecular mechanisms involved in EOC development. Ultimately, such tools will contribute to an accelerated diagnosis and better treatment of ovarian cancer.

According to the latest paradigm shift, tumors should be perceived as three-dimensional (3D) structures of high microenvironmental heterogeneity, including not only cellular but also acellular components. More precisely, the TME is a complex ecosystem that consists of the following: (a) multiple cell types, e.g., cancer associated fibroblasts (CAFs), endothelial cells, inflammatory cells, immune cells, adipocytes, mesenchymal stem cells, and pericytes; (b) vascularization; (c) the extracellular matrix (ECM) and (d) the secreted factors such as cytokines, chemokines and growth factors. All of those elements play an active role in cancer pathophysiology [4]. Particularly, fibroblasts are the most abundant cell type in the TME [5,6]. Primary fibroblasts and CAFs play a key role in in vitro tumor models as they contribute to the formation of a tumor microenvironment that supports tumor growth and progression. They secrete growth factors such as Vascular Endothelial Growth Factor (VEGF), Transforming Growth Factor β (TGF-β), Fibroblast Growth Factor (FGF), and Hepatocyte Growth Factor (HGF), which stimulate the proliferation of cancer cells and the development of blood vessels (angiogenesis), supplying oxygen and nutrients to the tumor model [7]. Additionally, CAFs actively create and remodel the extracellular matrix, which facilitates cancer cell invasion and metastasis [8]. Furthermore, fibroblasts modulate the immune response by secreting cytokines and chemokines that recruit immunosuppressive cells and suppress the body’s defense mechanisms against cancer cells [9]. Fibroblasts can also influence tumor resistance to anticancer drugs and induce an apoptosis-resistant phenotype in cancer cells [10].

Cellular monolayers are widely used in ovarian cancer research regarding combination therapies [11,12], chemoresistance mechanisms [13,14], and potential biomarkers [15,16]. Nevertheless, they do not faithfully reproduce the in vivo and pathophysiological processes. To bridge the gap between cellular monolayers and the in vivo conditions, various 3D models in vitro have been developed, including cell aggregates, spheroids, hydrogel models, scaffold cultures, and cell multilayers [17,18]. Such models may contribute to faithful mimicking of tumor phenotypic and genotypic features. They may also include the appropriate diffusion gradients of nutrients and oxygen [18,19]. However, a significant challenge remains in the development of cell models that mimic the heterogeneous and vascularized structure of gynecological tumor tissue, including non-cancerous elements of the tumor microenvironment (TME), such as blood vessels and fibroblasts, which play a critical role in tumor progression. One potential solution to this need is the use of Lab-on-a-Chip (LoC) and Organ-on-a-Chip (OoC) systems, which allow for the modeling of complex multicellular interactions under physical conditions resembling the in vivo environment. On-chip technologies enable control over key parameters, such as oxygen, nutrients, and drug gradients, which are difficult to achieve in standard 2D cultures [19].

Gynecological cancers, such as epithelial ovarian cancer, are characterized by significant heterogeneity. Within a tumor, different cell populations with distinct genetic, metabolic, and proliferative characteristics may coexist [20,21]. One of the key advantages of OoC systems is the ability to conduct multilayered cell cultures, which offer significant benefits over other 3D tumor models [22]. Multilayered cell cultures at the microscale allow for the recreation of various tumor regions with differing microenvironments (e.g., well- and poorly vascularized areas or regions with varying fibroblast densities). This enables research on gynecological cancers to resemble real in vivo conditions more closely, leading to a better evaluation of therapeutic efficacy. In gynecological tumor models, it is also crucial to replicate the process of angiogenesis—the formation of new blood vessels [23,24,25]. Tumors promote angiogenesis by secreting factors such as VEGF. Mimicking vascularization in in vitro models of gynecological tumors is essential for realistically simulating tumor growth, angiogenesis, hypoxia, drug transport, and metastasis [26,27]. The creation of vascular networks could help us understand many cell functions and mechanisms of various diseases. Perfusion microsystems, thanks to the supply of oxygen and nutrients as well as the removal of waste products from the cells, provide the necessary conditions to obtain vascularized microtissue. Therefore, microfluidic systems are ideal techniques for in vitro vascularization studies. So far, basic research based on the formation and characterization of vascular networks as well as tumor (based on hydrogel or spheroids) vascularization and angiogenesis has been studied in LoC systems [28,29,30]. In turn, multilayer culture in the microsystem will mimic the heterogeneous morphology of gynecological cancer tissue. There are several examples describing the research on interactions between non-malignant and cancer cells, but they were performed on non-multilayered cultures. Sung et al. presented a 3D microsystem for a non-multilayered co-culture of Breast Cancer Ductal Carcinoma In Situ cells (MCF-DCIS) with Human Mammary Fibroblasts (HMF) [31]. The presence of fibroblasts was found to promote the progression and invasion of cancer cells [32]. Although significant studies have been performed towards tissue vascularization and related vascularized tumor models, i.e., hydrogel culture of gynecological cells, self-organizing microvessel with spheroids and 3D gynecological cell culture [33,34,35], many challenges remain and should be solved.

One of the important challenges that deserves particular attention is the problem related to the quantitative determination of the specific contribution of each cell type to tumor progression. Although co-culture systems in newly developed OoC models increasingly allow the observation of complex interactions within the TME, they often do not enable selective analysis of how individual cell populations, such as fibroblasts or endothelial cells, influence the molecular or functional changes in the entire system. Advanced strategies, including targeted cell depletion, cell type-specific inhibition, or genetic labeling, are increasingly used to analyze these complex interactions in microsystems. For example, Yuan et al. used cell depletion in their study to perform a rapid evaluation of hepatotoxicity profiles of drug candidates using a designed biomimetic immune-liver-on-a-chip platform [36]. The integration of similar cell-specific analytical strategies with multicellular ovarian cancer-on-a-chip models could allow a more precise assessment of how stromal or vascular components influence tumor growth, inflammation, or angiogenesis. Such approaches may also further increase the mechanistic insight of studies on the ovarian cancer microenvironment and help to better understand the dynamic and heterogeneous nature of EOC progression.

In this report, we proposed an entirely novel approach to modeling EOC in vitro using microfluidic ToC systems. We developed an advanced microsystem enabling three-dimensional culture of ovarian cancer cells in co-culture with other cell types present in the TME: fibroblasts and vascular cells (Figure 1A). In contrast to other studies [23,37], the geometry and operating principle of our microfluidic system were based on the use of a poly(dimethylsiloxane) (PDMS) membrane as a structural element mimicking the function of the basement membrane underlying the epithelial layer of ovarian cells. Additionally, the multilayered 3D cell culture on the membrane was integrated with a hydrogel channel that served as a scaffold for creating a cylindrical blood vessel model. The multicellular model presented in this report allowed for the investigation of interactions between cancer cells and TME cells, including replication of the initial stages of angiogenesis, differentiation of primary cells into CAFs, examining processes of extracellular matrix remodeling that facilitate tumor cell invasion and migration, and detecting changes in gene expression indicative of inflammatory processes associated with EOC progression. This unique multicellular integration within a single system enabled, for the first time, the laboratory reconstruction of a 3D, heterogeneous, and vascularized EOC model.

2. Materials and Methods

2.1. Cell Culture

Human Umbilical Vein Endothelial Cells (HUVECs, Lonza, Visp, Switzerland) and Red Fluorescent Protein Human Umbilical Vein Endothelial Cells (RFP-HUVECs, PromoCell, Heidelberg, Germany) were cultured in Endothelial Cell Growth Medium-2 (EGM-2, PromoCell, C-22211, Germany) at 37 °C and 5% CO₂ up to 8 passages. Human Ovary Fibroblasts (HOFs) were sourced from ScienCell Research Laboratories. They were cultured in Fibroblast Medium (FM, ScienCell, 2301, Carlsbad, CA, USA) at 37 °C and 5% CO₂ up to 15 passages. HUVECs and HOFs were centrifuged at 1000 RPMI for 5 min during passaging to remove trypsin solution (Biowest, L0931, Nuaillé, France). Ovarian carcinoma cells (A2780) were sourced from Merck. They were cultured in Roswell Park Memorial Institute 1640 Medium (RPMI 1640, Biowest, L0500, France) at 37 °C and 5% CO₂. A2780 cells were centrifuged at 2000 RPMI for 3 min during passage to remove the trypsin solution.

2.2. Development and Fabrication of a Microsystem for Cell Culture

The microsystem consists of two layers made of PDMS separated by a thin PDMS membrane. PDMS layers were made using a resin mold fabricated using 3D printing technology (Profluidics 285D 3D printer, Concord, ON, Canada) using Master Mold resin (Cadworks3D, Concord, ON, Canada) according to the producer’s instructions. A resin mold was filled with non-cross-linked PDMS prepolymer solution mixed with a cross-linking agent in a weight ratio of 10:1 (Sylgard 184, Dow Corning, Midland, MI, USA). The prepolymer mixture was thoroughly degassed in a desiccator. Then, the mold was incubated at 85 °C for 30 min. Next, the PDMS cast was removed from a mold, rinsed with distilled water, and allowed to dry completely. A PDMS membrane was fabricated with the use of the same non-cross-linked PDMS prepolymer solution mixed with a cross-linking agent in a weight ratio of 10:1. The PDMS mixture was poured on the surface of a small Petri dish (6 cm in diameter) placed inside a spin-coater (1000 RPM, 60 s, Laurell Technologies, Lansdale, PA, USA). Then, a Petri dish was incubated at 65 °C for 3 h. After this time, rectangular membranes with appropriate dimensions (4 mm × 14 mm) were cut out with the use of a rectangular aluminum cutter. The membranes were cut out by applying the cutter to the surface of the PDMS and pressing on the press. Finally, one of the PDMS layers was bonded with a PDMS membrane using an oxygen plasma generator (time: 30 s, power: 80%, Diener). Then, the plasma activation process was repeated for both PDMS layers (with and without the membrane), and they were precisely connected. The developed microsystems were left under load for 24 h to ensure efficient bonding.

2.3. Lumen Formation

Lumens were created in the vascular channel (Figure 1B) using the Viscous Finger Patterning (VFP) technique [38]. First, the surface of the channels was modified to ensure good adhesion of the collagen matrix. More precisely, channels were filled with 2 mg/mL of polydopamine (Merck, Rahway, NJ, USA) in Tris-HCl buffer with a pH of 8.5 (Merck, USA). The microsystems were then incubated at room temperature (RT) for 1 h. After the incubation, the polydopamine solution was washed out by miliQ water. Then, microsystems were dried using compressed air and incubated for at least 15 min at 65 °C to ensure full evaporation of water. Next, the microsystems were cooled for 5 min at −18 °C. Simultaneously, collagen type I solution with the final concentration of 5 mg/mL and pH = 7 ÷ 7.5 was prepared by mixing phosphate-buffered saline solution (PBS) (1×) (Merck, USA), PBS (10×) (Gibco, Brooklyn, NY, USA), 1 M NaOH (Merck, USA) and rat tail collagen type I (Corning, Corning, NY, USA). Volumes of reagents were determined according to the producer’s instructions. A cold and well-mixed solution of collagen type I was introduced into a vascular channel using an automatic pipette until the channel was filled. Then, a single droplet of PBS 1× with a volume of 20 µL was placed on the surface of the outlet, which is a hole with a greater diameter (1.5 mm) [39]. Next, small droplets with a volume of 2 µL were introduced into the inlet of the channel. Consequently, PBS flow occurred. Therefore, a cylindrical channel surrounded by a thin layer of collagen matrix was formed. The microsystems were incubated at 37 °C for minimum 30 min to ensure hydrogel cross-linking. Then, a collagen type I with a final concentration of 0.1 mg/mL was introduced into cancer and stromal channels, both its upper and lower parts separated by a thin PDMS membrane. Finally, microsystems were placed in an incubator (37 °C, 5% CO₂) for at least 30 min. After that procedure, the microsystems were ready for cell seeding. The whole procedure was performed under sterile conditions and using sterile reagents.

2.4. Cell Loading and Their Culture in the Microsystem

All channels were carefully filled with appropriate media using a pipette. Then, a HUVEC suspension with a density of 10⁷ cells/mL was introduced into the vascular channel (Figure 1B). Next, a suspension of HOF cells with a density of 5 × 10⁶ cells/mL was prepared and loaded into the stromal channel. Microsystems were then placed on a droplet of EGM-2 media in a Petri dish. Holes in microsystems were facing downwards. Petri dishes containing microsystems were placed in an incubator (37 °C, 5% CO₂) for 3 h. After this time, the cell seeding procedure was performed again, but instead of HOF, A2780 suspensions with a density of 5 × 10⁶ cells/mL were loaded into the cancer channel (Figure 1B). This time, microsystems were placed in a Petri dish with the holes facing up. Each hole was covered with a drop of medium dedicated to the cells cultured in a given channel. A smaller dish filled with PBS (1×) was also placed in a large dish, minimizing the risk of liquid evaporation. A Petri dish containing microsystems was then placed in an incubator (37 °C, 5% CO₂). The channels were rinsed once daily with an appropriate medium using an automatic pipette. The analysis of the spatial arrangement of cells and their morphology was studied with the use of fluorescence microscopy (Olympus, Tokyo, Japan, IX71) at given time points (day 1, 3, 5, 7, 10). Additionally, the cells in culture were regularly flushed with fresh medium (appropriate for each of the cultured cell types) at the same time points. Flow in the channels of the microsystem was generated exclusively using an automatic pipette, and therefore parameters such as average flow velocity and shear stress were neither determined nor controlled. The flows generated in the system were transient and short-lived; therefore, quantitative parameters of fluid dynamics in the channels were not specified.

2.5. Viability Assay

The viability of the cells was analyzed based on differential staining using calcein acetoxymethyl ester (Calcein-AM, Merck, USA) and Propidium Iodide (PI, Merck, USA). A solution containing 1 μL of 1 mg/mL PI (aqueous solution), 1 μL of 2 mM CAM (dimethyl sulfoxide solution), and 500 μL of RPMI culture medium was introduced into the microsystem with cultured cells. The cells were incubated with fluorescent dyes for 10 min (37 °C, 5% CO₂), and they were observed with an inverted fluorescence microscope (Olympus IX71).

2.6. Cell Migration Analysis and Real-Time Observation

The migration study of HUVECs within connecting microchannels was conducted using a device designed for real-time tracking of fluorescently labeled cells (CytoSMART Lux3 Axion Biosystems, Atlanta, GA, USA). This experiment used HUVEC-RFP (PelloBiotech) cells exhibiting red fluorescence. A2780 and HOF cells were labeled with the green fluorescent dye CellTracker™ Green CMTPX (Invitrogen, Carlsbad, CA, USA). For this purpose, 1 mL of 5 µg/mL dye solution was added to the cells in a culture flask. The solution with the cells was incubated for 45 min (37 °C, 5% CO₂), after which the cells were detached and introduced into the microsystem following the previously described procedure. The migration study was conducted over 60 h, with time-lapse microscopic images captured every 15 min. Real-time cell growth observation studies could not be conducted for the full duration of the experiment in the microsystem (10 days) because of the technical limitations of the incubator-mounted camera and the inability to perform medium exchanges over such a long period, which would have resulted in air introduction into the microsystem. The distance traveled by HUVEC-RFP cells was measured after 12 h, 24 h, 36 h, 48 h, and 60 h. The average migration distance in a single connecting microchannel was used to calculate cell migration at each time point. Measurements were performed using ImageJ 1.54n software.

2.7. Immunostaining and Microscopic Analysis

Immunostaining of actin filaments of cells or α-smooth muscle actin (α-SMA) protein in the microsystems was carried out as follows. Firstly, all the microchannels were rinsed with PBS (1×), and 4% formaldehyde (Merck, USA) was introduced into them. Microsystems were then incubated for 30 min at RT. After the incubation, channels were washed with PBS (1×), filled with PBS (1×) containing 0.1% Triton X-100 (Merck, USA), and incubated for 20 min at RT. After that time, channels were rinsed with PBS (1×), filled with PBS (1×) containing 1% BSA and incubated for 1 h at RT. Afterward, channels were filled with a solution of 1% BSA (Merck, USA) containing antibodies. For immunostaining of actin filaments, it was Phalloidin conjugated with a fluorophore Alexa Fluor 568 (1:400, Invitrogen, A12380, USA). Next, microsystems were incubated for 1 h at RT and washed out with PBS (1×), leaving PBS solution within channels. For immunostaining the α-SMA protein, the Alpha-Smooth Muscle Actin Monoclonal Antibody (1A4) (0.5 mg/mL, 14-9760-82, Invitrogen, USA) it was used as a primary antibody. Next, the microsystems were incubated for 24 h at 4 °C. After that, the microchannels were washed three times with PBS solution (1×), and then a solution of secondary antibodies labeled with Alexa Fluor 488 (Goat anti-Mouse IgG, IgM (H + L) Secondary Antibody, 2 mg/mL, A10680, Invitrogen, USA) was introduced. The cells were incubated for 45 min at RT. Microscopic observation was performed using a confocal scanning microscope (Olympus FluoView FV10i or Zeiss Axio Observer 7 with LSM 900, Jena, Germany). Changes in fluorescence intensity of stained proteins were analyzed using ZEN 3.6 (blue edition) software. For each image, the total fluorescence intensity corresponding to the signal of fluorescently labeled α-SMA protein was determined. The analysis involved measuring the integrated fluorescence intensity within a defined region of interest (ROI) encompassing the entire field of view. The software automatically generated absolute numerical values of total fluorescence intensity, which were subsequently used for statistical analysis. For each experimental condition, at least n = 3 independent fields of view were analyzed. Data are presented as mean ± SD.

2.8. ELISA Assay

An ELISA assay was used to determine changes in the level of secreted Vascular Endothelial Growth Factor. For that purpose, microsystems were prepared in four variants. Variant 1 involved culturing only HUVECs in the lumen. Variant 2 involved culturing HUVEC and A2780 cells in the cancer channel. Variant 3 cultured HUVEC and HOF cells in the stromal channel. Variant 4 was the co-culture of HUVEC, A2780, and HOF in the respective channels. To obtain samples for the assay, the medium was collected from above the cells (about 20 µL), excluding HUVEC cells (vascular channel). In the control, the sample was taken from the middle microchannel, which was then filled with RPMI medium due to the lack of effect on VEGF secretion. Next, the microchannels were filled with fresh medium. The collected samples were transferred to an Eppendorf tube and stored at −20 °C. After collecting the samples, the microsystems were set aside in an incubator. Samples were collected at the same time at points 1, 3, 5, 7, and 10 days of culture. An immunoenzymatic assay was performed using the VEGF Human ELISA Kit (Invitrogen, USA) according to the manufacturer’s instructions.

2.9. Quantitative Real-Time Polymerase Chain Reaction (qRT-PCR) Analysis

A collagenase type I solution (Merck) at a concentration of 1 mg/mL was introduced into the vascular channel to isolate HUVECs from the collagen matrix. The microsystem with collagenase was incubated for 30 min (37 °C, 5% CO₂), and then trypsin solution (Biowest) was introduced into the remaining channels of the microsystem. After detaching all cells in the microsystem, the suspensions were collected and centrifuged (3 min, 2000 rpm). RNA was isolated from all cells cultured in the microsystem using the Total RNA Mini isolation kit (AA Biotechnology, Gdańsk, Poland), specially adapted for the isolation of RNA from small numbers of cells (<10⁶ cells/mL). Complementary DNA (cDNA) was synthesized by reverse transcription with a RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Scientific, Waltham, MA, USA). qRT-PCR was performed using SYBR^® Green Master Mix (Bio-Rad, Hercules, CA, USA) in the CFX Opus Real-Time PCR Systems (Bio-Rad). The primer sequences are listed in Table S1 (see Supplementary Materials). The relative expression levels were determined using the ΔΔCt method and were normalized to the expression of the housekeeping gene GAPDH, selected for its stable expression under our experimental conditions. To minimize variability related to cell number or confluency, equal amounts of total RNA were used for reverse transcription, and identical amounts of resulting cDNA were applied in each qPCR reaction. Results were collected and analyzed using CFX Maestro Software 2.3.

2.10. Statistical Analysis

At least three independent experiments were performed for each measurement (n ≥ 3). The obtained results were averaged, and the standard deviation (SD) was determined. In addition, a one-way analysis of variance was performed using the ANOVA test. Statistical significance was determined by a p-value of less than 0.05 (*).

3. Results

3.1. Characteristics of a Microsystem for Modeling Epithelial Ovarian Cancer (EOC)

A novel ToC microsystem was developed to replicate the functional co-culture of multiple types of ovarian cancer cells and their microenvironment under laboratory conditions. The aim of the first stage of the research was to develop a tool (chip) technologically adapted for (a) 3D lumen structure formed in hydrogel matrix for vascular formation, (b) 3D multilayer cell culture using layers mimicking basement membranes, and (c) studying vascular cell migration to simulate vascularization and/or angiogenesis processes accompanying ovarian cancer cell progression in vivo.

A microsystem for cell culture was developed with geometry comprising three main microchannels: a cancer channel for culturing ovarian cancer cells (A2780), a stromal channel for culturing human primary ovarian fibroblasts (HOF), and a vascular channel for culturing human umbilical vein endothelial cells (HUVEC) in a hydrogel matrix (Figure 1B). The tumor channel was located above the stromal channel, with both channels being 1000 µm wide and separated by a thin membrane made of poly(dimethylsiloxane) (PDMS) via spin-coating. The PDMS membrane was 50–56 µm thick and enabled co-culture of normal and cancer ovarian cells in a multilayered arrangement (on both sides of the membrane). Thanks to the properties of the membrane material, such as high porosity, biocompatibility, and permeability to small molecules [40], cells cultured on both sides of the membrane remained in contact. The membrane also served as a scaffold for ovarian cells growing on it and mimicked the function of the basement membrane found in vivo (Figure 1A). The vascular channel, 600 µm wide, was connected to the other channels through connecting microchannels (Figure 1B–D). The vascular channel was designed to integrate with the hydrogel matrix and reflect the structure of a cylindrical microvessel. Meanwhile, the connecting microchannels located between the multilayer cell structure and the microvessel were designed to allow the sprouting and migration of HUVECs from the vessel walls towards the tumor model in response to pro-angiogenic factors. Each of the three connecting microchannels was 500 µm long (Figure 1C,D).

3.2. Development and Analysis of 3D Cell Models in the Microsystem

The developed microsystem was used for the parallel culture of vascular cells and ovarian cells (co-culture of fibroblasts and cancer cells). In the first stage of the research, the microsystem was prepared for cell culture by filling the vascular channel with a hydrogel matrix (collagen type I), in which a cylindrical lumen was subsequently formed using the Viscous Finger Patterning method (see Section 2). In the next step, the same protein was used to coat the membrane on both sides to enhance the membrane’s hydrophilicity, improve ovarian cell adhesion, and facilitate cell migration and the exchange of secreted factors between all channels. After modifying the surface of the channels, cells were seeded into the microsystem in two steps: first, A2780 cells were introduced into the tumor channel, while the first part of the HUVEC cell suspension was simultaneously added to the lower part of the lumen. Next, HOF cells were introduced into the stromal channel along with the second part of the HUVEC cell suspension into the upper part of the lumen. The microsystem was inverted to position the cells in the appropriate regions of the membrane and lumen (Figure 1F). This approach produced an advanced, three-dimensional, vascularized, heterogeneous (multicellular) ovarian cancer model, which was subsequently characterized by shape, dimensions, and viability.

Over a 10-day culture period, a 3D, elongated structure composed of vascular endothelial cells was formed in the vascular channel filled with the collagen matrix. Analysis using Z-axis confocal imaging confirmed the cylindrical shape of the obtained cellular structure. Differential staining of HUVECs in the long-term culture demonstrated that the cells maintained a high level of viability (approximately 100%) throughout the experiment (Figure 2A). The average diameter of the microvessel models in the developed microsystems was 370 ± 9 µm (Figure 2C), indicating that the structures formed in the vascular channel closely resembled the shape and dimensions of small blood vessels (veins) in the human body [41,42]. Based on the average dimensions of the microvessels, their perfusion capacity was calculated using the following formula:

V = π⋅r²⋅h,

where V is the perfusion volume of the microvessel (µL), r is the average radius of the microvessel (~185 µm), and h is the average length of the microvessel (~12.5 mm).

The calculated perfusion capacity of 1.34 µL confirms that the microvessel volume is similar to the volume of small veins in vivo, which confirms that the developed model is scalable and suitable for future functional studies, e.g., on drug or nutrient transport [41,42]. Although no therapeutic agents were administered through the vessels in the present study, given the planned future applications, we performed an assessment of microvessel integrity by evaluating the permeability of a fluorescein solution. The microvessel integrity was confirmed, and the results are presented in the Supplementary Materials (Figure S3).

Although the average vessel diameter was relatively large (~370 µm), this dimension does not preclude structural and functional maturity. According to previous reports, vessels exceeding 300 µm can exhibit stable and mature characteristics when accompanied by the recruitment of mural cells, basement membrane deposition, and limited endothelial proliferation [43,44,45,46]. Moreover, similar vessel dimensions have been described in in vitro vascular models that displayed functional perfusion and pericyte coverage, confirming that stability can be achieved independently of vessel size [47]. Thus, the observed morphology and integrity of the obtained microvessels indicate that the developed constructs can be considered structurally mature and potentially suitable for long-term physiological studies.

To confirm the hypothesis that the microsystem could support the development of a vascularized model of epithelial ovarian cancer, the shape, dimensions, viability, and distribution of cells in the culture channels were analyzed for the complex multicellular culture. Microscopic analysis of the longitudinal and cross-sectional views of the cellular model demonstrated the successful formation of a 3D multilayer on the membrane, composed of cancer cells (A2780) and ovarian fibroblasts (HOF) (Figure 2B and Figure S1). The average thickness of the heterogeneous multilayer ovarian cell model on the 10th day of culture was 189 ± 17 µm (Figure 2C). The geometry of the microsystem was designed so that the membrane with the ovarian cell multilayer was positioned at the midpoint of the vascular channel. This arrangement, along with the presence of connecting microchannels, enabled direct contact between the endothelial and ovarian cells during the 10-day culture period. Microscopic images confirmed the presence of cells in the connecting microchannels. This observation may indicate active migration of the cells comprising the tumor model and suggest the activation of processes associated with angiogenesis or vascularization, which are characteristic of similar biological systems in vivo. Differential staining of all cell types in the vascularized ovarian cancer model unequivocally demonstrated its high viability (approx. 98–99%) (Figure 2B).

3.3. Tracking of Angiogenesis Processes in an Ovarian Cancer Model

In the developed microsystem, endothelial cells form a continuous monolayer lining the reconstructed vascular lumen, while the adjacent connecting microchannels are filled with collagen hydrogel. Although endothelial cells are capable of invading a hydrogel matrix, such migration does not occur spontaneously and requires chemotactic stimulation. In our previous work using an analogous microfluidic configuration lacking tumor (A2780) and stromal (HOF) cells, endothelial sprouting toward inter-pillar spaces was evaluated under controlled VEGF gradients. Importantly, migration into the connecting microchannels was not observed in the absence of chemoattractant stimulation [39]. The localization of cells in the connecting microchannels confirmed that cells cultured in the microsystem approach each other during the culture, and the obtained cellular structures (microvessel and cell multilayer) can merge into one model. This observation suggests that in the developed tumor model, mechanisms that stimulate the formation of new blood vessels because of the secretion of angiogenic factors, such as VEGF (vascular endothelial growth factor), were activated. These factors stimulate existing blood vessels to develop new microvessels that penetrate the tumor tissue and meet the increased demand for oxygen and nutrients, thus supporting tumor progression and determining the formation of metastases [48]. To confirm the active mechanism of angiogenesis in the developed cell model, real-time imaging of cellular changes and cell movement (migration) in the connecting microchannels was performed. For this purpose, HUVECs with the RFP gene encoding a protein exhibiting red fluorescence were used. The remaining cells in the tumor channel (A2780) and the stromal channel (HOF) were stained with Cell Tracker exhibiting green fluorescence. Within 60 h, significant changes in the position of vascular cells were observed, and their intensive migration towards the ovarian cell multilayer was proven. The distance covered by the cells during the observations was greater than 200 µm (Figure 3). The ability of HUVECs to migrate such distances indicates that soluble factors and cell–cell interactions provided by ovarian cancer cells and fibroblasts are sufficient to induce strong angiogenic responses. HUVEC cell migration in the presented model is consistent with the mechanisms of angiogenesis in vivo, where endothelial cells migrate towards a gradient of VEGF and other angiogenic factors, forming new blood vessels within the tumor [49,50]. The extent of migration further illustrates that the reconstructed vascular lumen is not merely structural but functionally responsive, mimicking the dynamic angiogenic behavior seen during tumor vascularization.

In addition to biochemical signaling, the physical properties of the extracellular matrix may also influence the observed angiogenic behavior. The stiffness of the collagen matrix (with concentration 5 mg/mL) used in the microsystem is known to support endothelial cell migration and sprouting without impairing lumen formation. Previous studies have demonstrated that matrices of moderate stiffness (0.2–1 kPa) promote angiogenesis, whereas excessively rigid matrices can inhibit endothelial invasion and capillary morphogenesis [51,52]. Therefore, the collagen matrix applied in our model likely provided a mechanically permissive environment conducive to vascular network formation.

3.4. Evaluation of the Functionality of the Developed Cell Models in the Microsystem

Preliminary observations suggested that angiogenic processes were active in the developed tumor model, accompanied by changes in the secretion of proangiogenic factors and other molecular-level alterations. To demonstrate the functionality of the EOC model in the microsystem, a biochemical and genetic characterization of the cells in the tumor model was performed. Changes in the expression of selected genes and the level of human VEGF were analyzed over consecutive days of culture, considering four cell culture variants differing in composition: (i) monoculture of HUVECs, (ii) co-culture of HUVEC cells with ovarian fibroblasts (HOF), (iii) co-culture of vascular cells with cancer cells (A2780), and (iv) a model comprising all studied cell types.

Based on immunoenzymatic analysis of hVEGF protein, it was unequivocally confirmed that at every experimental point, the most intense hVEGF secretion occurred in the most complex cellular culture variant. It was also confirmed that the level of secreted proangiogenic factor increased over time, reaching its maximum value (over 5000 pg/mL) on the last day of the experiment. The hVEGF level on day 10 of the culture was over 4800 pg/mL higher than on the first day, which served as the reference point (control) in this study (Figure 4A). The increasing secretion of hVEGF drove intensive vascular cell migration and confirmed the activation of angiogenic processes in the functional model of epithelial ovarian cancer.

As part of the study evaluating the functionality of the developed cellular model, the expression of three genes: angiogenin (ANG), angiopoietin-2 (ANGPT2), and interleukin-6 (IL-6) was analyzed. Angiogenin is a gene encoding a protein involved in the processes of new blood vessel growth and exhibits the ability to stimulate angiogenesis. Its mechanism of action is based on binding to receptors on the surface of vascular endothelial cells, activating signaling pathways that initiate their migration and proliferation [53]. A significant increase in angiogenin expression was observed on the third day of cell culture in the microsystem, which can be interpreted as the initiation of angiogenesis-related processes in the developed in vitro microsystem (Figure 4B). Angiopoietin-2 encodes a protein also involved in angiogenic processes that bind to endothelial cell receptors. However, it is responsible for destabilizing and increasing the permeability of existing blood vessels, thereby facilitating angiogenesis and the formation of new vessels [54]. In the case of this gene, a significant increase in expression was also observed, this time on the last day of culture (Figure 4C), which may suggest that angiogenic processes in the model may intensify during extended culture.

The expression of interleukin-6, a gene encoding a cytokine crucial in inflammatory responses, was also analyzed. IL-6 stimulates the production of acute-phase proteins, such as C-reactive protein (CRP), and is hyperactive in the tumor microenvironment [55,56]. A gradual increase in the expression of this gene was observed during the multicellular culture. The highest IL-6 expression level was recorded on the last day of the culture (Figure 4D). High expression of IL-6 indicates the presence of an inflammatory process in the developed tumor model. This implies that the model mimics a natural and functional TME, where proinflammatory cytokines enhance tumor progression and protect cancer cells from apoptosis. Additionally, IL-6 protein is one of the key mediators of the TME, and IL-6 gene elevated expression in the in vitro cellular model suggests that stromal cells (fibroblasts) cooperate with cancer cells to promote tumor growth and invasiveness. It is hypothesized that primary fibroblasts may have initiated differentiation into cancer-associated fibroblasts (CAFs), which can influence ECM remodeling [57]. Over 10 days, a significant transformation was observed, transitioning from evenly distributed cellular aggregates on the membrane to a densely packed structure resembling a tissue fragment (Figure S2).

3.5. Study of ECM Remodeling

α-SMA (alpha-smooth muscle actin) is an isoform of actin characteristic of smooth muscle cells. In the context of ovarian cancer, it plays a significant role in the TME and serves as a marker associated with tissue remodeling processes and cancer progression. α-SMA is a marker of cancer-associated fibroblasts (CAFs), responsible for the characteristic remodeling of the ECM in tumor tissue. Additionally, CAFs expressing α-SMA support angiogenesis and the migration of endothelial cells. Fibroblasts with high α-SMA expression can secrete growth factors (e.g., TGF-β, VEGF) that activate signaling pathways promoting cancer cell survival and resistance to treatment [58,59]. High α-SMA expression in the ovarian tumor microenvironment is associated with more aggressive disease progression and greater metastatic potential. In terms of diagnostics and therapy, α-SMA is an important research target as a potential prognostic and therapeutic marker. Therapies aimed at modifying fibroblast activity may limit tumor progression. Therefore, in this study, changes in the levels of α-SMA protein were also analyzed in the developed cellular model, based on fluorescence intensity analysis of the protein after immunostaining. The studies presented in Figure 5 show that the fluorescence intensity associated with the stained α-SMA increased with each day of culture. This is evidence of a change in the phenotype of HOFs to CAFs. The increase in the amount of the α-SMA marker enhances the remodeling of the ECM present in the model and activates the angiogenesis process, which confirms the observed migration of HUVECs (Figure 3).

4. Discussion

Ovarian cancer is reported to be the third most common gynecological cancer and the eighth leading cause of death among women worldwide. Ovarian cancer is a disease with a survival rate of around 15% at five years. This is due to the lack of visible and characteristic symptoms at the onset of the disease, which leads to more than 80% of patients being diagnosed at an advanced stage of the disease [1,2,3]. To understand the pathophysiology of ovarian cancer and enhance the ability to screen drugs, it is necessary to develop appropriate in vitro models that reflect the complexity of the ovarian cancer microenvironment. In this context, recent advances in 3D cell culture and microfluidics offer new opportunities to develop highly innovative models that could in the future (i) understand the pathophysiology of ovarian cancer, (ii) accelerate the screening process at the preclinical stage and (iii) increase the probability of correctly evaluating potential anticancer drugs.

Although in vitro studies aimed at elucidating the mechanism of cancer development or evaluating different types of drugs in cancer models are still performed on 2D cultures, more advanced models, such as 3D models or the OoC approach, are increasingly being used. Also, in the case of ovarian cancer modeling, there are many examples of research conducted on 3D models. However, these are mainly studies conducted on aggregate/spheroid models. For example, Tofani et al. described the generation and characterization of 3D multicellular spheroids using ovarian cancer cells (SKOV-3) in co-culture with mesenchymal cells (MUC-9) or fibroblasts (CCD27-Sk) [60]. Wojtowicz et al. performed a detailed analysis of the effect of ovarian CAFs on the morphology and gene expression of A2780 and W1 cells cultured in both 2D and 3D as cell aggregates [61]. Baka et al. developed a 3D-printed model including ovarian cancer cells (SKOV-3) and cancer-associated fibroblasts (Me-Wo) encapsulated in a biocompatible gelatin-alginate hydrogel [5]. Moving forward, 3D tumor models in vitro may be incorporated within Organ-on-a-Chip technology becoming a promising alternative to conventional models. Importantly, such models may simulate the flow of physiological fluids in the body and the mapping of biochemical and biomechanical signaling between diverse cell types. Only a few examples of Ovarian-Cancer-on-Chip have been developed. Surendran et al. engineered a microfluidic platform to recreate an immunocompetent microenvironment modeling dynamic neutrophil migration and 3D invasion of ovarian tumor cells [62]. Li et al. described a microfluidic platform for the generation of ovarian cancer–mesothelial cells, where ovarian cancer spheroids were co-cultured with primary human peritoneal mesothelial cells [63]. Another example is the work of Rizvi et al., who designed a linear microfluidic system to study the effect of flow on ovarian cancer cell adhesion and growth [64]. However, referring to the EOC model aggregate/spheroid modeling does not represent the natural arrangement of cells. In simple terms, in the case of EOC, epithelial tumor cells form 3D structures on a stroma composed of extracellular matrix (ECM) and anchored fibroblasts activated into cancer-associated fibroblasts (CAFs) and blood vessels (Figure 1A). In this work, an attempt was made to model EOC in a way that closely resembles its architecture in vivo. For this purpose, a model of a cell multilayer was created, embedded on the collagen-coated PDMS membrane, where on one side there was a layer of fibroblasts (HOF) and on the other side there was a layer of ovarian cancer cells (A2780), connected by microchannels to a 3D blood microvessel model (HUVEC) embedded in a collagen matrix (Figure 1B). The advanced EOC model was cultured for 10 days, during which all cells were characterized by high viability (Figure 2). In our study, PDMS was selected as the material for the membrane supporting the multilayered cell culture because, although it exhibits limited permeability to larger molecules (e.g., proteins) [65,66], it still provides several key advantages for our application. The membrane is not the sole point of interaction between cells, as they also communicate indirectly through the connecting microchannels. This microsystem design ensures effective exchange of exogenous factors between cell types, supporting the dynamic behavior of the culture. Moreover, the PDMS membrane closely mimics the physical properties of the basement membrane, offering elasticity, transparency, and mechanical support for the cell layers. Its structural stability and optical clarity allow for the formation of robust, multilayered, and functional tissue-like constructs.

Ovarian tumor tissue accumulates various cell types. Cancer cells, together with cancer-associated fibroblasts (CAFs) constitute the main cell group that builds ovarian tumors. Cancer cells produce TGF-β, stimulating CAFs. CAFs are stromal cells that contribute to the synthesis and remodeling of ECM from a physiological to a pathological environment. This ability of CAFs enables pro-tumor activity promoting metastasis. The combination of ECM remodeling by CAFs and the secretion of pro-inflammatory molecules (cytokines, growth factors) is responsible for the diverse effects of the mechanical properties of the matrix on tumor growth, invasion, and, concomitantly, angiogenesis [67]. In the previously described works on the creation of in vitro cell models of ovarian cancer, both monocultures [68,69] and co-cultures [60,61,70] of ovarian cancer models were used. In our studies, to better reflect the in vivo microenvironment to generate an EOC model, we used several cell types to develop a model exhibiting features of heterogeneity. We used only a single ovarian cancer cell line (A2780), which does not fully reflect the subclonal heterogeneity or patient-to-patient variability observed in ovarian tumors in vivo. However, in the context of the present study, the term “heterogeneity” refers to the coexistence of multiple distinct cell types within a single model, including cancer cells, fibroblasts, fibroblasts differentiating toward CAFs, and endothelial cells, which together create a more complex and dynamic in vitro system compared with a monoculture of a single cell line. It is worth noting that most studies described in the literature often used fibroblasts other than those derived from the ovary, for example, skin fibroblasts [60,71]. HOF cells do not have a cancer-associated fibroblast phenotype. However, in the presented work, we showed that their co-culture with cancer cells activated them to CAFs (Figure 5). This was observed based on the increased levels of the extracellular matrix remodeling marker (α-SMA protein) in the cellular multilayer. Although our study did not include housekeeping protein as a loading control and therefore the analysis focused on detecting relative changes in α-SMA expression, the obtained results are sufficient to confirm the tendency of HOF cells to shift their phenotype toward CAFs. The change in phenotype was probably the result of HOF stimulation by TGF-β secreted by A2780 cancer cells. A similar correlation was described in the studies by Wu et al. [72] and Axemaker et al. [73]. Moreover, evidence that stromal cells (fibroblasts) cooperate with cancer cells to promote tumor growth and invasiveness is the observed increase in the expression of the gene encoding the IL-6 protein (Figure 4A).

As mentioned above, the ovarian tumor microenvironment, in addition to tumor cells and fibroblasts, includes endothelial cells. The endothelial cells are the crucial component of the ovarian TME. Endothelial cells are responsible for the formation of new capillary networks playing an essential role in tumor development providing the delivery of nutrients and oxygen to a growing tumor. Endothelial cells may also promote tumor development via paracrine pathways, and they do influence the effects of chemotherapy. In the context of ovarian cancer treatment, there are only a few reports of the in vitro co-culture models to study the cellular mechanisms of tumor-vasculature interplay and to evaluate the therapeutic efficacies of drugs in such complex in vivo microenvironments. Wan et al. proposed a 3D in vitro co-culture model for cancer angiogenesis studies to test drugs with anti-angiogenic potential [74]. Saha et al. developed a microfluidic device integrating an ovarian tumor model, vascular model, and blood. The identification of complex tumor-blood cell interactions and metastatic signaling for antiplatelet therapeutics was studied with the use of the created device [24]. Ando et al. developed a vascularized hypoxic tumor model which interfaces a microfluidic 3D microvasculature with a 3D hypoxic tumor model [75]. The EOC-on-Chip model we created considers the presence of endothelial cells and in particular maps their full architecture. In our work, we used the Viscous Finger Patterning method to create a 3D lumen in which we seeded HUVECs. A 3D vessel model was also created in the work of Saha et al. and Ando et al., but neither considered fibroblasts, which are extremely important in modeling colon cancer. However, our studies have shown that fibroblasts activated in CAFs have an extremely important effect on the angiogenesis process. We have demonstrated that the presence of activated fibroblasts is necessary for HUVEC migration toward the tumor-containing multilayer, as shown by both microscopic images (Figure 3) and the amount of VEGF (Figure 4D). Although VEGF can be released by tumor cells, fibroblasts are the primary source of host-derived VEGF. VEGF induces neovascularization and microvascular permeability, leading to the extravasation of plasma proteins such as fibrin, which then attracts fibroblasts, inflammatory cells, and endothelial cells. These cells produce ECM, causing desmoplasia, which re-enhances tumor angiogenesis. Although VEGF inhibition (e.g., with bevacizumab) was not employed in the present study, the observed endothelial migration and vessel formation are consistent with the VEGF-driven angiogenic mechanisms described in previous reports. Similar microvascular responses have been observed in studies where VEGF blockade served as a control condition. Therefore, the angiogenic behavior demonstrated in our system aligns well with established VEGF-dependent processes, supporting the validity of our model despite the absence of pharmacological inhibition. Furthermore, the formation of new vessels in tumors from pre-existing vessels is associated with the expression of certain genes, including those of the angiopoietin family. In this work, we showed that ANG and ANGPT2 gene expression increase over time in the EOC model, which is evidence for the observed migration of HUVEC cells towards the multilayer.

The development of a cylindrical, lumen-forming vascular structure within the microsystem represents an important achievement and allowed us to confirm barrier integrity through fluorescein-based permeability assays (Figure S3). Moreover, the determined value of the microvessel perfusion capacity confirms the similarity of the developed structure to in vivo conditions. Nevertheless, one limitation of the present work is that endothelial maturation and tight junction formation were not directly assessed by immunostaining of junctional proteins (e.g., ZO-1, VE-cadherin) or by TEER measurements. Addressing this aspect will be the focus of future studies to provide a more comprehensive validation of vascular barrier functionality in the proposed model.

In the presented results, an inverse temporal relationship between interleukin-6 and angiogenin expression was observed: a high level of ANG expression in the early stage of culture (day 3), followed by a decrease in subsequent days, accompanied by an increase in IL-6 expression, peaking on day 10. This relationship may appear contradictory to previous reports indicating that IL-6 can induce ANG expression and thereby promote angiogenesis [76]. However, several important factors should be considered that may influence the dynamics of gene expression in the context of a 3D in vitro model. First, IL-6 may act both as a promoter of angiogenesis and as a mediator of chronic inflammation, whose effect on cells can depend, among other things, on the duration of the inflammatory state. Early ANG expression may result from a rapid endothelial cell response to the culture conditions and growth factors present during the initial stage of culture [77]. As the culture progresses and the microvascular-like structures become more organized, the demand for strong pro-angiogenic stimulation may decrease, resulting in lower ANG expression at later time points. In contrast, the increased expression of IL-6 at a later stage may be a consequence of the progressive activation of stromal and cancer cells, leading to a chronic inflammatory response and remodeling of the microenvironment [78]. Such chronic inflammation may not result in an immediate upregulation of ANG but rather lead to more complex, delayed responses, such as changes in cell phenotype or activation of alternative angiogenic pathways. The observed inverse temporal pattern of IL-6 and ANG expression may thus reflect a complex molecular regulatory network characteristic of an in vitro model that mimics the dynamic conditions of the TME.

Looking ahead, the presented model offers several avenues for refinement and expansion to further enhance its translational potential. One key direction will be adaptation of the platform for drug testing, including the evaluation of hypoxia- and EMT-associated markers, as well as quantification of angiogenic and inflammatory responses under treatment conditions. Such analyses will enable systematic comparison of drug sensitivity and resistance profiles with clinically observed patient data, thereby strengthening the model’s predictive value. In parallel, the system will be engineered toward high-throughput applications, for example, through integration with plate readers, incorporation of electrodes for real-time monitoring, and implementation of TEER measurements. This would allow dynamic and quantitative assessment of vascular integrity while maintaining a culture timeframe short enough to preserve the structural and functional complexity of the tumor microenvironment. In addition, the current system does not yet incorporate immune components, which represent critical drivers of tumor progression. For instance, tumor-associated macrophages (TAMs) are known to promote angiogenesis, modulate extracellular matrix remodeling, and suppress anti-tumor immune responses, thus representing a key element of the ovarian cancer microenvironment. Future iterations of the model could therefore be expanded to include TAMs, T cells, or natural killer cells, allowing interrogation of tumor–immune interactions and their contribution to therapy resistance. The modular design of the platform also enables the integration of further stromal or vascular elements, which may improve its capacity to capture the dynamic remodeling of tumor vasculature over extended periods.

Together, these future directions highlight the flexibility of the presented platform and its potential to evolve into a robust tool for studying ovarian cancer biology, tumor-stroma-immune interactions, and for supporting preclinical drug screening in a physiologically relevant context.

5. Conclusions

A new microsystem was successfully developed to model the complex structure of epithelial ovarian cancer. The developed cellular model contained the two most important features of epithelial ovarian cancer: microenvironmental heterogeneity and vascularization. The new cell model was evaluated for functionality. It has been shown that: (a) HUVECs cultured in the form of microvessels in a collagen matrix successfully migrate towards the multilayer model of cancer tissue, (b) a complex multicellular model of ovarian cancer secreted pro-angiogenic factors, (c) in long-term in vitro culture, inflammation (characteristic of tumor tissue) may occur. The developed model will be able to be used in drug screening, testing the effectiveness of anticancer therapies, and in personalized medicine.