Short cold atmospheric plasma treatment preserves vascular quiescence in 3D tumor-stroma-endothelial models of pancreatic cancer in vitro and in ovo Authors: Ruben Verloy, Emma Peeters, Angela Privat-Maldonado, Sophie Rovers, Ho Wa Lau, Louize Brants, Christophe Hermans, Jorrit De Waele, Christophe Deben, Evelien Smits, Annemie Bogaerts Published: 2025 ABSTRACT AbstractPancreatic ductal adenocarcinoma (PDAC) features a dense desmoplastic stroma, limiting drug delivery and promoting chemoresistance. While anti-angiogenic strategies have shown limited success, underexplored pro-angiogenic approaches may improve perfusion and treatment efficacy. Cold atmospheric plasma (CAP) generates reactive oxygen and nitrogen species and demonstrates anti-cancer effects at longer treatments and pro-healing, angiogenic effects at shorter treatments. We investigated the impact of short CAP treatment on angiogenesis in PDAC using a triple co-culture spheroid model and a turkey in ovo model, mimicking the tumor microenvironment. CAP was applied via the kINPen MED for 15–60 s in vitro and 10–30 s in ovo. In vitro, CAP inhibited endothelial tube formation treatment time-dependently and reduced VEGF-A secretion, while other angiogenic factors remained unchanged. In ovo, vascularization around the tumor was slightly increased in the BH tumors, while not significantly altered within the tumors, except for a significant reduction after 30 s of CAP treatment. Additionally, a significant increase in tumor weight was observed following short CAP treatment. These findings suggest that under standard conditions, short CAP treatment preserves vascular quiescence in PDAC and may exert subtle tumor-modulating effects, highlighting the importance of treatment parameters and model complexity in CAP-based therapeutic strategies.Subject terms: Gastrointestinal cancer, Cancer microenvironment Introduction Pancreatic ductal adenocarcinoma (PDAC) is projected to become the second leading cause of cancer-related death by 2030, highlighting the urgent need for advancements in treatment strategies1. Representing ~85% of all pancreatic cancer cases, PDAC is marked by a dense, fibrous tumor stroma caused by a desmoplastic reaction2. This reaction causes high intratumoral pressure, forming a physical barrier that compresses blood vessels, and reduces vascularity, thereby limiting drug delivery and fostering chemoresistance3. Endothelial cells (ECs), beyond their crucial role in angiogenesis, could add to the complexity of PDAC by transitioning into cancer-associated fibroblasts (CAFs) through an endothelial-to-mesenchymal transition, exacerbating the disease4,5. Additionally, activated pancreatic stellate cells (PSCs), a specific fibroblast subtype, significantly contribute to desmoplasia and modulating the tumor microenvironment (TME)6. This complex milieu promotes PDAC traits such as rapid growth, high invasiveness and metastatic potential, survival under hypoxic and nutrient-poor conditions, and immunosuppression6. Significant attention in cancer research has been given to anti-angiogenic approaches to inhibit vascularity and tumor growth by starving the tumor. However, these approaches have struggled to maintain their promising pre-clinical effects when transitioning to clinical models7–10. In the context of PDAC, such strategies often exacerbate tumor aggressiveness and reduce the efficacy of chemotherapeutics by hindering drug delivery and inducing hypoxia11. Conversely, pro-angiogenic approaches for PDAC have been proposed to enhance the delivery and effectiveness of chemotherapeutics, demonstrating reductions in tumor growth, metastasis, and desmoplasia, alongside increases in survival rates, blood vessel density, and perfusion12,13. Cold atmospheric plasma (CAP) is a partially ionized gas that consists of reactive oxygen and nitrogen species (RONS), which induces oxidative stress on cells and operates at low temperatures (<40 °C). CAP treatment received increasing attention in anti-cancer research for its ability to selectively kill cancer cells14,15. On the other hand, CAP also shows promising results in wound healing, which suggests it can promote angiogenesis, proliferation and tissue oxygenation at lower treatment intensities, consistent with the hormetic response induced by CAP16–22. Low-intensity CAP treatment can induce angiogenesis, with nitric oxide (NO) and other RONS like hydrogen peroxide, superoxide, and hydroxyl radicals playing key roles in the angiogenesis pathway22–24. This controlled oxidative stress, or oxidative eustress, activates cell proliferation pathways23. ECs, crucial for new blood vessel formation, respond to CAP by expressing pro-angiogenic factors such as FGF-2 and angiopoietin-2, encouraging EC proliferation and migration18,23,24. Paracrine signaling between fibroblasts and ECs further enhances this activation18. Additionally, RONS disrupt EC membranes, releasing more FGF-2 and stimulating neighboring EC proliferation24. Based on information from the manufacturer of the kINPen MED, 30 to 60 s/cm2 promotes tissue regeneration and is therefore seen as short CAP treatment25–27. In addition, 60 s CAP treatment of 2-dimensional (2D) PDAC cell lines, 3-dimensional (3D) spheroids and in ovo PDAC tumors already showed significant tumor cell death28,29. Thus, controlled RONS delivery via CAP could regulate blood vessel formation in tumors, offering a promising approach to modulate angiogenesis in cancer. In the present study, we investigate whether pro-angiogenic effects observed in wound healing by short CAP treatment can be achieved in cancer models to enhance chemotherapy delivery and efficacy against PDAC. For this, we have used our in-house developed triple co-culture (TCC) spheroid models, which incorporate cancer cells, PSCs, and ECs30. Our in vitro TCC spheroid models overcome the limitations of 2D cultures and classic monoculture 3D spheroids by including different cell types to better mimic the tissue architecture and more accurately resemble cell-cell and cell-matrix interactions31–33, better capturing the complexity of the TME31. In addition, we used the turkey in ovo model, a variant of the popular chick chorioallantoic membrane (CAM) model, which is ideal for angiogenesis studies due to its visible angiogenic network and longer embryonic development period, which allows for a longer experimental window without the need for an ethical approval34,35. Results CAP treatment does not induce tube formation in vitro First, we evaluated the angiogenic potential of CAP treatment on ECs. For this, we performed the tube formation assay using spheroid-conditioned medium collected three days after CAP treatment (indirect effect due to a change in secreted cytokines after CAP treatment) (Fig. 1a). We observe that CAP treatment has a treatment time-dependent inhibitory effect, as it consistently reduced the number of branching points, number of tubes, covered area, and total tube length (Fig. 1b). TCC spheroids with the MiaPaCa-2 cell line treated for 45 and 60 s presented the most significant differences (p < 0.05). The heatmaps reveal varying responses to CAP treatment across the different spheroid types, each representing a unique TME (Fig. 1c). The shift from green to red indicates a general reduction in the measured angiogenic parameters with increasing treatment time, suggesting CAP treatment inhibits angiogenesis in these spheroid models. However, the extent of these changes varies among the spheroid types, e.g., with BxPC-3:RLT-PSC:HMEC-1 spheroid-conditioned medium. No change in tube formation upon CAP treatment was observed, highlighting PDAC heterogeneity. This suggests that different tumor cell populations or microenvironments respond differently to the treatment, emphasizing the need to consider intra-tumor variability in therapeutic strategies. CAP treatment inhibits tube formation in TCC spheroids. a Methodology scheme of using CAP-treated spheroid-conditioned medium for a tube formation assay to evaluate angiogenesis. b Two sets of representative images of the tube formation analysis with raw images and processed images on the left and right, respectively. c Heat map of the IKOSA tube formation analysis results of four different spheroid types, showing the effects of CAP treatment per quantitative parameter. Data is represented as mean (n ≥ 9 from three independent experiments). Statistics were performed using linear mixed model and Dunnett’s test with the treatment as fixed effect using JMP Pro software after the removal of significant outliers (GraphPad Software v10.1.0). *p ≤ 0.05. Overall, these results indicate a treatment time dependent inhibitory effect of CAP treatment on the angiogenic potential of ECs to form tubes. Angiogenic cytokines do not change after CAP treatment To investigate whether this subtle inhibition of angiogenesis was related to the secretion of angiogenic growth factors by the treated PCCs, PSCs and ECs, a MesoScale Discovery assay was performed (Fig. 2a). All analytes measured represent circulating soluble proteins in the supernatant. Tie-2, although a receptor tyrosine kinase, was measured in its soluble form (sTie-2), as detected by the MSD assay. While no statistically significant differences are observed in the secretion of VEGF-D, b-FGF, PlGF, and VEGF-A across different CAP treatment times and TCC spheroid types, several trends are noted (Fig. 2b-e). VEGF-A concentrations decrease with increasing treatment times, suggesting reduced secretion of this angiogenic growth factor under prolonged treatment (Fig. 2b). In addition, the BxPC-3:hPSC21:HMEC-1 spheroids have a notably higher secretion of VEGF-A compared to other spheroid types. VEGF-D data does not show a clear trend (Fig. 2c). In contrast, b-FGF and PlGF secretions remain relatively stable across the different treatment times, with MiaPaCa-2 TCC spheroids secreting more b-FGF than BxPC-3 spheroids and MiaPaCa-2:hPSC21:HMEC-1 spheroids secreting lower amounts of PlGF (Fig. 2d, e). Tie-2, Flt-1 and VEGF-C were not detected in the samples. Although these trends are not statistically robust, they suggest that the inhibitory effects of CAP treatment on tube formation could be due to a reduction in VEGF-A in the CAP-treated spheroid-conditioned media. Revealing the secretion profile of angiogenic growth factors in the TCC spheroids upon CAP treatment. a Methodology scheme of using CAP-treated spheroid-conditioned medium for a MesoScale Discovery assay of seven soluble growth factors: VEGF-A, VEGF-C, VEGF-D, Tie-2, Flt-1, PlGF and bFGF. Bar plots of the secretion levels of b VEGF-A, c VEGF-D, d bFGF, e PlGF in TCCspheroid-conditioned medium. Data is represented as mean ± SEM (n = 2–3 from three independent experiments after removal of outliers) with each dot representing one independent experiment of 4-6 pooled TCC spheroids of the same condition. UT = untreated; VEGF-A/C/D = Vascular endothelial growth factor A/C/D; Tie-2 = Angiopoietin-1 reception; Flt-1 = Vascular endothelial growth factor receptor 1; PlGF = Placental growth factor; bFGF = Basic fibroblast growth factor. Statistics were performed in GraphPad using one-way Anova or Kruskal-Wallis test after the removal of significant outliers (GraphPad Software v10.1.0). CAP treatment induces a tumor-specific angiogenic response around the tumor in ovo To investigate the effect of CAP treatment on tumor angiogenesis in a more complex model, we used the turkey in ovo model, as it has an extensive and visible vascular network. We evaluated three short CAP treatment times (10 s, 20 s and 30 s of CAP), as 45 s and 60 s treatments showed signs of angiogenesis inhibition in vitro. To account for the heterogeneity of PDAC, we seeded two different co-cultures: MiaPaCa-2 with RLT-PSC (MR tumors) and BxPC-3 with hPSC21 (BH tumors). Macroscopic angiogenesis was evaluated by looking at the vascular network surrounding the tumor using the IKOSA tool (Fig. 3a, b), which provides quantitative measures of the vasculature around the tumor, including the total vessel length, the total vessel area, the mean vessel thickness and the number of branching points (Figs. 3c and 2S). Capturing the effect of CAP treatment on the macroscopic angiogenesis in the turkey in ovo model using the IKOSA analysis. a Methodology scheme of tumor formation, CAP treatment and IKOSA analysis using two different PCC and PSC lines. b A representative raw image (left) and the IKOSA analysis masking (right) are shown. The tumor is indicated with a green circle. c Heat map of the IKOSA analysis results of two different tumor types, showing the effects of CAP treatment on the total vessel length, total vessel area, mean vessel thickness and the number of branching points. Data is represented as mean (n ≥ 7 from four independent experiments). UT = untreated. Statistics were performed in GraphPad using one-way Anova or Kruskal-Wallis test after the removal of significant outliers (GraphPad Software v10.1.0). *p ≤ 0.05. For all four parameters, only one statistically significant difference is observed between the three different treatment groups (10 s, 20 s and 30 s of CAP) and the untreated group for both tumor types (Figs. 3c and 2S). Within the BH tumors, 10 s and 20 s CAP treatment shows a trend to increase the total vessel length, total vessel area and the number of branching points, with a significant increase in the 20 s CAP treatment group for the mean vessel thickness (p < 0.05). Within the MR group, 10 s of CAP treatment showed an increase in branching point, however, no clear trend is observed in general. Overall, these results indicate that the vasculature around the tumor remains mostly unchanged upon short CAP treatment, with minor induction of angiogenesis in the BH group. Moreover, the varying responses among tumor types highlight the importance of accounting for tumor heterogeneity in CAP research (Fig. 3c). CAP treatment does not induce angiogenesis within the tumor in ovo Next, angiogenesis was evaluated within the PDAC tumors of the turkey in ovo model by collecting them three days after CAP treatment, paraffin-embedding and staining them with H&E to differentiate the turkey red blood cells from the other human and turkey cells (Fig. 4a). Pictures were taken of the stained tissue and analyzed with QuPath to quantify the percentage of turkey red blood cells within the tumor tissue, which indicate the presence of blood vessels (Fig. 4b). For both tumor types, all three treatment groups show a trend to decrease the percentage of red blood cells compared to the untreated group. Except for a significant decrease in the percentage of red blood cells after 30 s of CAP treatment in the BH tumor type, small to no differences are observed in the percentage of red blood cells present in both tumor types treated with CAP for different times (Fig. 4c). Capturing the effect of CAP on microscopic angiogenesis and tumor weights in ovo. a Methodology scheme of tumor formation, CAP treatment and measuring the number of turkey red blood cells 3 days after treatment. Two different PCC and PSC lines were used. b QuPath analysis of the H&E stained tissue sample. The human tumor cells stained intensely blue/purple with hematoxylin and the turkey red blood cells stained intensely pink (left). The turkey red blood cells are indicated in blue after QuPath analysis (green boxes). c Bar plots of the QuPath H&E analysis results, showing the percentage of red blood cells compared to the total number of cells in the tumor. Data is represented as mean ± SD (n ≥ 6 from three independent experiments). d This data is showing the tumor weights per treatment group. Data is represented as mean ± SD (n ≥ 16 from five independent experiments). BH = BxPC-3 + hPSC21 tumor; MR = MiaPaCa-2 + RLT-PSC tumor; UT = untreated. Each dot represents one egg. Statistics were performed in GraphPad using one-way ANOVA or Kruskal-Wallis test after the removal of significant outliers (GraphPad Software v10.1.0). *p ≤ 0.05; **p ≤ 0.01. Besides looking at the induced angiogenesis around and within the in ovo tumors, we measured the weight of the tumors on the day of harvesting (Fig. 4d). For both BH and MR tumors, all CAP-treated groups show a higher mean tumor weight compared to the untreated groups, except the 30 s CAP treatment MR group. Furthermore, all three CAP treatments seem to significantly increase tumor weight of BH tumors, compared to the untreated group (Fig. 4d, p < 0.05 or p < 0.01). For the MR tumor type, 10 s CAP gives the highest mean tumor weight. Altogether, our in ovo data suggests that a short CAP treatment preserves vascular quiescence in PDAC with no consistent increase or decrease in angiogenesis around or within the tumor, while promoting tumor growth in the BH tumors (Figs. 3 and 4). Discussion PDAC remains one of the most aggressive cancer types due to its dense stroma, poor vascularity, and resistance to therapies. Limited, but effective, pro-angiogenic strategies aim to enhance vascular perfusion, potentially improving drug delivery and treatment outcomes12,13. However, systemic drugs have negative side effects as they induce an effect on the entire human body, which can be prevented using a local treatment, like CAP. CAP is known for its ability to generate RONS and has shown a pro-angiogenic potential, as shown in vitro, in ovo and in vivo with various devices18,21,22,24,36–39. We investigated whether CAP could promote angiogenesis in PDAC using in vitro TCC spheroid models and an in ovo turkey model. This could open opportunities to enhance the effects of chemotherapy in this hard-to-treat cancer. However, CAP treatment slightly inhibited endothelial tube formation in vitro in a treatment time dependent manner, and VEGF-A secretion in TCC spheroids trended downward with longer CAP exposure. Other cytokines, such as b-FGF and PlGF, remained stable, indicating that CAP-induced oxidative stress may disrupt pro-angiogenic signaling without triggering compensatory increases in growth factors. The observed effects in the tube formation assay could be due to the direct effects of CAP on our spheroids leading to a different growth factor secretion profile of these cells, indirect effects by remaining long-lived RONS in the medium, or the generation of secondary reactive species as a product of the reaction of CAP-derived RONS with organic molecules in the medium. However, RONS are quickly consumed in presence of organic molecules, like in blood, and their concentration would be negligible low 72 h post-treatment40. In addition, indirect effects due to changes in the medium are also of value, as this reflects a clinical situation where the tumor microenvironment is also influenced by CAP treatment41. Indeed, 60 s of CAP-treated nanoparticles have been associated with reduced VEGF mRNA expression in breast adenocarcinoma cells as well42. Similarly, direct CAP exposure (15 sec) using the kINPen MED reduced cell growth, metabolism, tube formation, and 2D migration in human dermal microvascular ECs, comparable in origin to our HMEC-1 line43. In contrast, short CAP treatment using a dielectric barrier discharge (DBD) device has been shown to enhance EC proliferation and increase FGF-2 release39. The MicroPlaSter ß plasma torch (30 s, 4 slm) also upregulated FGF-2, angiopoietin, and VEGF receptor expression in ECs, and FGF-2 in fibroblasts (2 min, 4 slm)18. Tube formation improved after 30 s of direct CAP treatment, and supernatants from CAP-treated fibroblasts (1–2 min) also promoted angiogenesis18, in contrast to our findings, where supernatants tended to suppress endothelial activity. Other studies using in-house plasma jets with helium (direct) or nitrogen (30–360 s indirect) showed enhanced tube formation, migration, and VEGF receptor activation, with mechanisms involving eNOS and possibly AMPK signaling37,38. Together, this suggests that inducing oxidative eustress at short CAP treatments may be difficult to achieve with the kINPen MED under standard conditions and highlight CAP’s dual-edged nature: its angiogenic effects vary widely depending on factors such as device type, gas composition, type of liquid, cancer and EC type, microenvironment, treatment mode and duration. Many of these factors critically influence not only the type but also the quantity of RONS generated, ultimately resulting in highly variable biological effects44. However, due to the complexity of this large number of variables, fully predicting the outcomes remains challenging. Next, we investigated CAP’s effect on angiogenesis using the in ovo model, as this model has an extensive vascular network, ideal to investigate angiogenesis. We used the turkey CAM model instead of the chicken, as the extended development time of the turkey embryo compared to the chicken provides a longer experimental time window that is beneficial for combining CAP with chemotherapy in future studies. Analysis of vascularization around the tumor revealed limited significant pro-angiogenic changes in the BH group and unchanged vascularization in the MR group. Vascularization within the tumor showed a slight inhibitory trend with increasing CAP treatment times in both groups with statistical significance in the BH group at 30 s of CAP treatment. In addition, an increase in tumor weight was observed with short CAP treatments in the BH group, suggesting CAP influences tumor growth, as expected due to hormesis at short CAP exposures, through mechanisms unrelated to angiogenesis45. In contrast, CAP treatments of 60 seconds have previously been associated with tumor shrinkage in pancreatic cancer using the kINPen device, indicating that the therapeutic outcome is indeed treatment time dependent28. This tumor-promoting effect underscores the need to better understand CAP’s broader biological impact, particularly its interaction with stromal components and cellular survival pathways. In contrast, other studies have reported pro-angiogenic effects under different conditions and in non-cancer contexts. One study using a DBD device showed that plasma enhanced EC proliferation, migration, and tube formation via ROS-induced FGF-2 release24. Similarly, daily CAP treatment of mouse skin wounds over 10 days increased CD31 and FGF-2 mRNA expression18. In a separate in ovo study using the kINPen MED, all treatment durations between 30 and 300 seconds increased vessel area, especially after several days21. However, this study evaluated angiogenesis in a healthy CAM model, without tumor tissue. Additionally, it used plasma-treated PBS (indirect treatment), lacking short-lived species like NO, supporting the idea that longer-lived species such as H₂O₂ may play a more prominent role in promoting angiogenesis, while short-lived species may be more detrimental. Both studies also relied on repeated treatments, which may be necessary to drive angiogenic responses. In our case, a single treatment was chosen to align with the clinical context of PDAC, where repeated CAP applications are currently not feasible. This distinction may explain the absence of pro-angiogenic effects in our model and highlights the challenge of overcoming the inherent oxidative stress and stromal complexity of PDAC. These findings emphasize the importance of using complex, disease-relevant models and staying mindful of clinical translation when assessing CAP’s therapeutic potential. Future pro-angiogenic CAP research should optimize CAP parameters (a different plasma device, e.g. the MicroPlaSter ß plasma torch, as mentioned above, or changing the feed gas for the kINPen MED to helium, or argon with admixtures of O2 and N2 as this might lead to higher concentrations of relevant RONS for angiogenesis, such as nitric oxide, hydroxyl radical, hydrogen peroxide)37,46–50. Another future perspective is to further explore CAP’s mechanistic effects on the tumor-stroma-vasculature interplay and investigate its potential synergistic use with chemotherapy. While CAP’s ability to selectively target cancer cells remains promising, its limited pro-angiogenic effects and potential tumor-promoting effects with short treatments in PDAC emphasize the need for careful optimization before integration into multimodal therapeutic strategies. Using CAP treatment for PDAC patients could be particularly beneficial for resectable, borderline resectable and locally advanced unresectable patients that respond well to chemotherapy and are eligible for partial surgical resection51. In these patients, relapse often occurs due to metastatic sites formed from leftover tumor cells52. Current surgical challenges, such as the need to leave residual tissues around major arteries for safety, lead to high recurrence rates and poor survival outcomes53. By treating the surgical margins with CAP, residual cancer cells can be killed, which is already investigated in other cancer types in a phase I clinical study54, or can be killed better by chemotherapy upon angiogenesis induction. Pro-angiogenic therapies have been proposed as an innovative strategy against PDAC, and limited research of CAP has shown pro-angiogenic effects on wound healing or endothelial cells. However, our research indicates under the conditions used here, that short kINPen treatments cannot overcome PDAC’s resistance against vascular modulation. This negative result helps refining the therapeutic scope of CAP, indicating that its application in PDAC may need to be combined with other modalities, or further optimization of CAP parameters is required to achieve microenvironmental reprogramming. Moreover, it underscores the importance of context-specific evaluation of CAP effects, as its impact may differ across tumor types and treatment parameters. To conclude, CAP demonstrated no pro-angiogenic effects in vitro, with treatment time dependent inhibition of endothelial tube formation and reduced VEGF-A secretion. In the in ovo model, CAP preserved vascular quiescence both around and within the PDAC tumors. However, in the BH group, a slight increase in vascularization around the tumor was observed. Additionally, short CAP treatments were associated with an increase in tumor weight, suggesting potential tumor-promoting effects that may be independent of angiogenesis. While the kINPen MED at short treatment times and argon as feed gas was not able to induce angiogenesis in PDAC, other experimental setups (e.g. a different feed gas, such as helium or argon with admixtures of O2 and/or N2, or a different device, such as the MicroPlaSter ß plasma torch or a DBD device) could be explored to achieve a pro-angiogenic response in PDAC with short CAP treatment. These findings emphasize the complexity of modulating PDAC’s TME and the need for more research using CAP treatment as a novel therapy. Future research should explore its mechanistic effects and combination with other therapies to maximize therapeutic outcomes. Methods Cell culture The human pancreatic cancer cell (PCC) lines MiaPaCa-2 (ATCC, Manassas, VA, USA) and BxPC-3 (ATCC, Manassas, VA, USA) were used in this study. The human immortalized PSC line RLT-PSC was kindly provided by Prof. Ralf Jesenofsky, of the Faculty of Medicine, University of Mannheim55, while the hPSC21 line was established at Tohoku University, Graduate School of Medicine, kindly provided by Prof. Atsushi Masamune56. In addition, the immortalized EC line HMEC-1 (ATCC, Manassas, VA, USA) was used. All cell lines were used for a maximum of 20 passages after thawing, tested for mycoplasma every 3 months, and STR profiling has validated that the cell lines remained identical to the original cell line. Cancer and stellate cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, 10938025, Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS, 10270106, Life Technologies, Carlsbad, CA, USA), 100 U/mL penicillin, 100 μg/mL streptomycin (15140122, Life Technologies, Carlsbad, CA, USA), and 2 mM L-glutamine (25030024, Life Technologies, Carlsbad, CA, USA). Cancer and stellate cells were transduced with NucLight Rapid Red (4741, Essen Biosciences, Ann Arbor, MI, USA) to express mKate2 and NucLight Green (4475, Essen Biosciences, Ann Arbor, MI, USA) to express GFP, respectively. HMEC-1 cells were cultured in MCDB131 (10372019, Thermo Fisher Scientific, Waltham MA, USA) supplemented with 10 ng/mL Epidermal Growth Factor (PHG0314, Thermo Fisher Scientific, Waltham, MA, USA), 1 µg/mL Hydrocortisone (H0396, Sigma Aldrich, Saint Louis, MO, USA), 10 mM L-glutamine (25030024, Life Technologies, Carlsbad, CA, USA), 10% FBS (10270106, Life Technologies, Carlsbad, CA, USA), and 100 U/mL penicillin, 100 μg/mL streptomycin (15140122, Life Technologies, Carlsbad, CA, USA). Cells were maintained in a humidified incubator at 37 °C and 5% CO2. Spheroid formation Four different TCC spheroids were formed as previously described30. Briefly, spheroids were seeded at 7000 (BxPC-3:hPSC21:HMEC-1 in a ratio of 6:5:3) or 3500 (MiaPaCa-2:RLT-PSC:HMEC-1 in a ratio of 6:3:3) cells per well, including 0.24% methylcellulose in ultra-low adherent (ULA) plates (7007, Corning, Corning, NY, USA). Empty outer wells were filled with water to minimize evaporation and plates were sealed with a breathable membrane (Z380059, Merck, Rahway, NJ, USA). The ULA plates were centrifuged at 453 × g for 10 min. Turkey chorioallantoic membrane (CAM) model The turkey eggs (Broeierij Claeys, Kruisem, Belgium) were introduced into the laboratory at day six of their embryonic development and kept in the dark at 19–21 °C overnight. The next day, after checking fertility with a flashlight, fertilized eggs were placed in an incubator at 37.7 °C with 65% humidity, cooled for two hours daily, and turned (lateral rotation from left to right) hourly to prevent adhesion and embryonic death. On day seven, eggs were disinfected with 70% isopropanol quick pads (QS0029, MLS, Menen, Belgium), punctured with a 20 G needle (613-3937, VWR, Radnor, PA, USA) on the small side of the egg to relocate the air pocket, and sealed with Leukosilk tape (46834-00 AP, Essity, Stockholm, Sweden) before being incubated upright without cooling and turning. On day nine, PCCs and PSCs were seeded onto the CAM. After shell opening, a filter paper soaked in diethyl ether was applied to a large vessel to erode it, and a silicon ring was placed on top of the vessel. A two million cell pellet (2:1 ratio of PCC:PSC) was resuspended in 20 µL Matrigel (734-0269, VWR, Radnor, PA, USA) and was allowed to solidify as a droplet hanging from the tip for ~1 min before seeding within the ring. Eggs were sealed with Tegaderm (16002, 3 M, Saint Paul, MN, USA) and placed back in the incubator for four days. On day thirteen, the tumors received CAP treatment. By day sixteen, tumors were harvested. The eggs were removed from the incubator, Tegaderm was taken off, and 750 µL paraformaldehyde was applied to fix the vascularized membrane. The egg was incised along the midline to flip the CAM and tumor into a weighing boat for photographing. The embryos were sacrificed by decapitation, and the tumors were dissected, weighed, and embedded in paraffin for hematoxylin and eosin (H&E) staining. All experiments were conducted in accordance with the University of Antwerp guidelines. Following the Flemish Laboratory Animal Committee and European Union guidelines, the study did not require the animal ethics committee approval, as the experimentation did not exceed two thirds of embryonic development (ED19). Cold atmospheric plasma (CAP) treatment The kINPen MED (adjusted to operate below 4 slm) was used to treat the spheroids and in ovo tissues. The kINPen MED is a plasma jet device that generates an atmospheric cold plasma beam below 40 °C. Prior to CAP treatment of spheroids, 150 μL of medium was removed and 50 μL of fresh MCDB131 medium was added. Argon was used as ionizing gas at 2 standard liter per min (slm) to treat both spheroids and in ovo tissues in the turkey egg. The treatment was performed for 15, 30, 45 or 60 s, with a 12 mm gap in MCDB131 medium (spheroids) or 10, 20, 30 s, based on the in vitro results, and a 10 mm gap (in ovo tissues) between the kINPen MED nozzle and the liquid surface. The plasma setup was controlled by an automated stage using the program WinPC-NC. Spheroid-conditioned medium was collected 72 h post-treatment for the tube formation assay and MSD analysis. Tube formation assay Assessment of angiogenesis was done with a tube formation assay in a 384-well plate (3764, Corning, NY, USA) as described before30. Briefly, three-day-old TCC spheroid supernatants (200 μL) were refreshed by removing 150 μL and adding 50 μL (final 100 μL), corresponding to a 50% refresh. Wells were coated with 8 μL Cultrex (3533, Bio-Techne, Minneapolis, MN, USA), centrifuged (180 × g, 3 min), and incubated 1 h at 37 °C and 5% CO₂. HMEC-1 cells (58,000 cells/mL) were seeded in 40 μL of either MCDB131 or CAP-treated spheroid-conditioned medium. After 6 h, images were acquired with a 4× objective (Spark Cyto, Tecan) and cropped using Python v3.12 to remove meniscus-related edges. Tube formation parameters (loops, covered area, tube length, branch points) were quantified via IKOSA (Kolaido, Altenrheim, Switzerland). MSD The Mesoscale Discovery (MSD) assay for angiogenesis was performed using the MSD V-PLEX Angiogenesis Panel 1 Kit (Catalogue No. K15190G-1, Rahway, NJ, USA) and an MSD MULTI-SPOT 96-well plate. The assay was performed following manufacturer’s instructions. Results are presented in duplicates or triplicates where each replicate is a pool of 4–6 spheroids of an independent experiment, totaling an average of 8–18 spheroids per condition. Hematoxylin and eosin staining The 5 µm thick sections were cut from paraffin embedded blocks containing the PDAC tumor tissue. These sections were baked on microscope slides for 2 h at 60 °C, ensuring secure attachment of the tissue to the slide surface. Next, the slides were dewaxed by washing in 100% xylene for 10 min, 100% isopropanol for 6 min, 95% isopropanol for 6 min, and 75% isopropanol for 6 min. Thereafter, H&E staining was performed with the following steps: 12 min incubation in hematoxylin, rinsing with water, 2 min incubation in sodium carbonate, rinsing with water, 5 min incubation in eosin, and rinsing with water. To finalize the staining process, slides were dehydrated with 75% isopropanol for 2 min, 95% isopropanol 2 min, 100% isopropanol for 2 min, and 100% xylene for 2 min. A coverslip with permanent mounting medium was carefully applied to the slide to preserve the stained tissue sections for microscopic examination. Statistical analysis All experiments were performed at least three times with three technical replicates, unless specified otherwise. Outlier tests were performed using GraphPad Software v10.1.0. Statistics for the tube formation assay were performed using linear mixed model with the treatment as fixed effect using JMP Pro v17.0.0 software. For multiple comparisons of the experimental groups to the untreated control, Dunnett’s test was used. To calculate the statistical significances between the concentrations of angiogenic growth factors in spheroid-conditioned media using the MesoScale Discovery assay and the in ovo experiments, all experimental groups were compared to their control group using the one-way ANOVA or Kruskal-Wallis test. Supplementary information Supplementary information Acknowledgements We acknowledge financial support from the Fund for Scientific Research (FWO) Flanders (Grant ID 1SD6522N, 1S85226N, G004524N), the Research Fund of UAntwerp (FFB210293) and Methusalem funding of the University of Antwerp. We would also like to thank several patrons, as part of this research was funded by donations from different donors, including Dedert Schilde vzw, Mr. Willy Floren, and the Vereycken family. We like to acknowledge the support from the European Cooperation in Science & Technology (COST) Action on “Therapeutical applications of Cold Plasmas” (CA20114; PlasTHER). All graphs and some statistical analyses were made and performed with GraphPad Prism. Other statistical analyses were performed with JMP. All figures were created with BioRender.com. Author contributions R.V. designed and performed the study, acquired and analyzed data, and drafted this article. E.P. performed the study, acquired and analyzed data, and drafted this article. A.P.M. contributed to the design and supervision of the study, and drafting of this article. S.R., H.W.L., L.B. and C.H. acquired data. CD provided resources, contributed to the methodology and drafting of this article. J.D.W. supported the visualization and drafting of this article. A.B. and E.S. co-supervised this study, provided resources, and contributed to the drafting of this article. All authors read and approved the final manuscript. Data availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Ruben Verloy, Emma Peeters. These authors jointly supervised this work: Evelien Smits, Annemie Bogaerts. Supplementary information The online version contains supplementary material available at 10.1038/s41698-025-01241-9. References Associated Data Supplementary Materials Supplementary information Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.