Introduction

Drug development is a lengthy, expensive and multi-step process, from therapeutic target identification and optimization to clinical approval. Furthermore, it is acknowledged that more than half of all drugs fail to reach Phase II or III of clinical trials1. In order to overcome this, it is essential to improve the predictive power of preclinical stages, while simultaneously promoting animal replacement in order to respect the Replacement, Reduction and Refinement (3R) rule2. In this context, cell culture modelling appears to be the most suitable alternative in the drug development pipeline. Indeed, cell culture is easy to handle, quite inexpensive and provides results rapidly3. Monolayer cell culture is classically regarded as the gold standard method due to its high degree of reproducibility and repeatability. However, this approach involves culturing cells on flat plastic supports which do not represent the in vivo cell environment. Consequently, the data obtained from these experiments are not sufficient for accurately predicting drug doses. To address this challenge, 3D cell cultures have been developed over the past decades, offering a mean to bridge the gap between conventional 2D in vitro cell line culture and in vivo experimentation. Firstly, this approach enables to more accurately mimic physiological responses. Since 3D models have a considerable degree of similarity with solid tissues, once they reach a diameter of 500 μm, their conformation enables the establishment of concentric layers, allowing the formation of gradients of oxygen, nutrients and metabolites, such as lactates4,5,6,7. Secondly, this permits the generation of a model that closely mimics the drug diffusion within cellular masses, enabling more accurate predictions regarding both the efficacy and the resistance of molecules3,7,8,9. 3D conformation promotes cell-cell and cell-matrix interactions that are absent in monolayer models, thus increasing the structural complexity and bringing the model closer to the in vivo environment. Moreover, it has been demonstrated that gene expression profiles of 3D models more accurately reflect those of clinical tissues than those of 2D cell cultures8. Additionally, they present an exponential growth kinetic, analogous to avascular tumors, followed by a plateau phase that closely resembles the initial phase of vascular growth5. It is also an easier method to apply before animal testing, reducing the number of animals used in preclinical phases3,4. In addition, it is possible to produce both pathological and healthy models, which is advantageous for assessing the toxicity of a drug and, at the same time, predict potential adverse effects. To date, there are lots of different techniques available from spontaneous aggregation to microfluidic microarray, including scaffold-based development or inserts4,5,10,11. Nevertheless, a significant limitation is the lack of standardization in these 3D culture processes, which results in very variable results and leads consequently to the impossibility of clearly identifying a lead molecule for clinical trials.

In this context, this work focused on the deployment of a simple, cost-effective, reproducible, standardized method of production and analysis of 3D cell culture, which remains viable for up to 14 days and could be used in any standard cell culture laboratory. The easily adaptable protocol is based on liquid overlay technique12,13,14 and has been improved and applied to six cell lines in order to obtain spheroids designed for Medium-Throughput Screening (MTS). We have designed a large panel composed of non-tumoral and tumor-derived models from different tissues. The tumor models comprise A-375 (melanoma, BRAF mutation15, NCI-H1975 (non-small cell lung cancer, EGFR mutation16, NCI-H1417 (small cell lung cancer), and SK-OV-3 (ovarian cancer), all selected for their well-characterized genetic backgrounds and clinical relevance. Additionally, the panel also includes a non-tumorigenic breast model (MCF-10A) to evaluate breast anticancer drug side effects, and a healthy kidney monkey tissue (VERO 76) widely used in virology and vaccine research. These models complement our previous breast cancer models (MDA-MB-231 and SUM1315) by expanding to additional lineages, enabling the assessment of drug responses across a wider range of genetic contexts and tissue origins. With the aim of analyzing treatment efficacy, we have characterized these models by assessing their structure, size, viability and metabolic activity. In order to confirm the relevance and the coherence of our panel, we have demonstrated that anti-Poly-ADP Ribose Polymerase (anti-PARP) treatments had the expected effect on BRCA1 (BReast-CAncer susceptibility gene) wild-type and mutated cancer models, as well as on non-tumorigenic tissue.

Results

Optimal conditions for 3D cell culture formation

In this study, the spheroids features were compared according to two parameters, i.e. cell concentration and matrix percentage, in order to select the best combination. For each cell line, four cell concentrations ranging from 200 to 10,000 cells per well were tested along with four matrix conditions ranging from 0% to 3% in a 96-well plate in culture medium. The cell and matrix concentrations were selected to obtain a single, size-controlled and viable spheroids based on the following criteria: (i) one single and round spheroid per well: based on bright field acquisition, because it allows the algorithm to more accurately measure the spheroid’s diameter, (ii) a spheroid with a minimum diameter of 500 μm on day 14: to permit a naked-eye detection and facilitate their handling, (iii) a proliferative spheroid: based on a potential diameter increase between days 1 and 14, to quantify the cytostatic and cytotoxic effects of drugs, (iv) a viable spheroid: based on the live/dead test acquisition, to assess the qualitative cytotoxic impact of drugs, and (v) a metabolically active spheroid: based on a resazurin test, to quantify the cytotoxic effect of the drugs. When identical results were obtained with different combinations of cells and matrix, the condition requiring a smaller concentration of cell was chosen.

Single spheroid formation per well

For the development of the melanoma (A-375) model, all conditions permitted the formation of spheroids (Fig. 1, a). However, conditions with 0% matrix percentage resulted in many disaggregated cells around the spheroid. Additionally, none of the cell conditions with 3% matrix concentration produced a single micro-tissue per well. In fact, there were many small satellite spheroids around the main one, which could interfere with the automated measurement of diameter. In contrast, 1% and 2% matrix concentrations allowed the formation of a single spheroid per well.

For the small cell lung cancer (NCI-H1417) model, the same results as for the melanoma model were obtained for the 0% matrix condition with disaggregated cells around the spheroid (Fig. 1, b). Then, concentrations of 5,000 and 10,000 cells per well led to the formation of satellite micro-spheroids, which are incompatible with the required characteristic of a single model per well. In conclusion, concentrations of 200 and 1,000 cells per well led to the formation of a single spheroid per well when formed with matrix.

Fig. 1
figure 1

Bright field acquisitions of 3D models at the end of the culture (day 14). (a) A-375 melanoma, (b) NCI-H1417 small cell lung cancer, (c) NCI-H1975 non-small cell lung cancer, (d) SK-OV-3 ovarian cancer, (e) MCF-10A non-tumorigenic breast and (f) VERO 76 kidney healthy cells were seeded at four cell concentrations, from 200 to 10,000 cells per well, with four matrix conditions, from 0% to 3%, in culture medium. To study their unicity, bright field acquisitions were made after 14 days of culture, using Cytation10MV cell analyzer (Agilent, M = 4X, scale bar = 500 μm).

Full size image

Then, for the other lung cancer (NCI-H1975) model, ovarian cancer (SK-OV-3) model, non-tumorigenic breast (MCF-10A) model, and kidney (VERO 76) model, all combinations of matrix concentrations (0% to 3%) and cell concentrations (200 to 10,000 cells per well) resulted in a single micro-tissue per well (Fig. 1, c–f).

Spheroid diameter

For the melanoma (A-375) model, all conditions resulted in a spheroid diameter superior to 500 μm at day 14 (Fig. 2, a). Additionally, as observed with bright-field acquisitions, the conditions without matrix presented disaggregated cells, which interfered with the algorithm. Consequently, the associated spheroids growth curves were misinterpreted. Indeed, the 0% matrix conditions appear to have a higher proliferation rate, whereas this only reflects the disaggregation of the spheroids. The day 14/day 1 (D14/D1) proliferation ratio for conditions containing matrix ranged from 0.97 ± 0.05 for the 10,000 cells per well/1% matrix combination to 3.55 ± 0.57 for the 200 cells per well/3% matrix combination. The 200 cells per well with matrix conditions had a significantly higher D14/D1 ratio with 3.52 ± 0.26 for 1% matrix, 3.24 ± 0.25 for 2% matrix, 3.55 ± 0.38 for 3% matrix, than 1,000 cells per well with 1.69 ± 0.12 for 1% matrix, 2.16 ± 0.12 for 2% matrix, and 2.25 ± 0.15 for 3% matrix (p < 0.001).

For the small cell lung cancer (NCI-H1417) model, the conditions with 200 cells per well did not reach the minimum diameter of 500 μm. On day 14, the sizes were 353.8 ± 11.8 μm for 0% matrix, 314.5 ± 22.2 μm for 1% matrix, 275.4 ± 11.9 μm for 2% matrix, and 270 ± 31.3 μm for 3% matrix (Fig. 2, b). Disaggregation of the spheroids without matrix also led to a biased spheroids growth curve, as observed with the melanoma model. The D14/D1 proliferation ratios for the conditions with matrix were significantly higher for 1,000 cells per well than for 5,000 (p < 0.001) or 10,000 cells per well (p < 0.001).

For the non-small cell lung cancer (NCH-1975) model, three conditions did not reach the required diameter: 200 cells per well with 0% matrix (224.4 ± 70.3 μm), 200 cells per well with 3% matrix (446.3 ± 22.0 μm) and 1,000 cells per well with 0% matrix (302.5 ± 62.9 μm) (Fig. 2, c). Conditions without matrix had a significantly lower D14/D1 ratio than conditions with 1% (p < 0.001), 2% (p = 0.002), or 3% (p = 0.004) matrix.

For the ovarian cancer (SK-OV-3) model, three conditions did not reach the required diameter: 200 cells per well with 0% matrix (367.7 ± 80.9 μm), 200 cells per well with 1% matrix (497.4 ± 68.3 μm) and 200 cells per well with 3% matrix (442.0 ± 48.0 μm) (Fig. 2, d). Regarding cell concentrations, the D14/D1 proliferation ratio for the conditions with matrix was significantly higher for 1,000 cells per well than for 5,000 (p < 0.001) or 10,000 cells per well (p < 0.001).

For the non-tumorigenic breast (MCF-10A) model, eight conditions did not reach the required diameter (Fig. 2, e). The spheroids diameter of all 200 cells per well concentrations was less than 500 μm, with an average value of 107.9 ± 4.8 μm for 0%, 333.8 ± 102.3 μm for 1%, 253.1 ± 56.5 μm for 2%, and 231.1 ± 55.6 μm for 3%. Additionally, four other cell/matrix combinations did not reach the minimum diameter: 1,000 cells per well with 0% matrix (180.8 ± 19.9 μm), 1,000 cells per well with 3% matrix (394.1 ± 105.5 μm), 5,000 cells per well with 0% matrix (304.1 ± 21.4 μm), and 10,000 cells per well with 0% matrix (360.2 ± 36.2 μm). Conditions without matrix had a significantly lower D14/D1 ratio than conditions with 1%, 2%, or 3% matrix (p < 0.001, for each condition). However, there was no significant difference among other matrix concentrations (p = 0.998 for the comparison of 1% and 2%, p = 0.808 for the comparison of 1% and 3%, and p = 0.942 for the comparison of 2% and 3%).

For the healthy kidney (VERO 76) model, six conditions did not reach the required diameter of 500 μm: 200 cells per well with 0% matrix (163.2 ± 27.3 μm), 200 cells per well with 1% matrix (381.9 ± 52.3 μm), 200 cells per well with 2% matrix (413.6 ± 20.5 μm), 200 cells per well with 3% matrix (437.7 ± 49.5 μm), the 1,000 cells per well with 0% matrix (276.0 ± 24.4 μm) and 5,000 cells per well with 0% matrix (424.1 ± 24.3 μm) (Fig. 2, f). Conditions without matrix had a significantly lower D14/D1 ratio than conditions with 1%, 2%, or 3% matrix (p = 0.0006, p < 0.001 and p < 0.001, respectively). Regarding cell concentrations, the D14/D1 proliferation ratio was significantly higher for 1,000 cells per well than for 5,000 (p = 0.001) or 10,000 cells per well (p < 0.001).

Spheroid viability

For the viability assay, the (A-375) melanoma spheroid model consisted of viable cells (shown in green) and dead cells (shown in red) (Fig. 3, a). The yellow regions indicate areas where red dead cells overlap with green live cells, suggesting a high cell density and intermixing. This pattern reflects the highly proliferative nature of the cells forming these spheroids and is consistent with previous observations. The high number of red dead cells in the spheroids could correspond to necrotic areas resulting from nutrient deprivation, as often observed in dense 3D cultures. The highest live/dead cell ratio was recorded in conditions seeded with 200 cells per well.

Fig. 2
figure 2

Growth kinetic curves of 3D cell cultures. (a) A-375 melanoma, (b) NCI-H1417 small cell lung cancer, (c) NCI-H1975 non-small cell lung cancer, (d) SK-OV-3 ovarian cancer, (e) MCF-10A non-tumorigenic breast and (f) VERO 76 kidney healthy cells were seeded at four cell concentrations, from 200 to 10,000 cells per well, with four matrix conditions, from 0% to 3%, in culture medium. To evaluate their growth kinetic, spheroid size was measured on different days until 14 days of culture, on bright field acquisitions with Gen 5.03 software (Agilent). Data represent means ± SD. A Kruskal-Wallis test has been performed.

Full size image

According to the previous analysis, complete disaggregation of the small cell lung cancer (NCI-H1417) spheroids was observed for 0% matrix concentration (Fig. 3, b). For the more proliferative cell conditions (1,000 cells per well with matrix), the viability profile appeared similar for all the matrix conditions.

The results were similar for the non-small cell lung cancer (NCI-H1975) 3D cell culture, with slightly higher viability for 1% matrix concentration with 1,000 to 10,000 cells per well (Fig. 3, c).

For the ovarian (SK-OV-3) model, no spheroids were detected for the 0% matrix condition. As a result, no data is available regarding viability and metabolic activity this condition. Spheroids with 200 and 1,000 cells per well exhibited greater viability compared to those seeded with 5,000 and 10,000 cells per well. Figures (3, d) illustrates this trend, showing the spheroids as yellow due to the overlap of green viable cells and red dead cells.

For the non-tumorigenic breast (MCF-10A) model, spheroid viability was lower for the two lower cell concentrations (200 and 1,000 cell per well) compared to the two higher concentrations (5,000 and 10,000 cells per well) (Fig. 3, e).

Finally, for the kidney (VERO 76) model, all conditions with matrix had a similar live/dead cell ratio (Fig. 3f).

Spheroid metabolic activity

Regarding the metabolic activity of the melanoma (A-375) model, all conditions (matrix percentage and cell concentration) resulted in similar values ranging from 19.2 ± 0.9 UF for conditions of 10,000 cells per well with 0% matrix, to 25.7 ± 3.1 UF for conditions of 10,000 cells per well with 2% matrix (Fig. 4, a). Two conditions were chosen based on their elevated D14/D1 size ratio. Among them, the combination of 200 cells per well with a 2% matrix yielded a significantly higher metabolic activity (25.1 ± 1.1 UF) compared to the same cell density with a 1% matrix (22.0 ± 1.1 UF, p = 0.021).

For the small cell lung cancer (NCI-H1417) model, the metabolic activity of the more proliferative cell conditions (1,000 cells per well with matrix) did not differ significantly, with values of 9.0 ± 0.9 UF for the 1% matrix, 9.2 ± 0.5 UF for the 2% matrix, and 10.3 ± 1.5 UF for the 3% matrix (p = 1.000 for the 1% and 2% comparison, p = 0.319 for the 1% and 3% comparison, and p = 0.481 for the 2% and 3% comparison) (Fig. 4, b). Consequently, these three conditions could be suitable for drug screening.

The metabolic activity for the non-small cell lung cancer (NCI-H1975) model did not differ significantly among the three most viable conditions. The values were 9.0 ± 0.9 UF for 1,000 cells per well and 1% matrix, 8.7 ± 1.3 UF for 5,000 cells per well with 1% matrix and 9.1 ± 0.8 UF for 10,000 cells per well with 3% matrix (p = 0.974 for the 1,000 and 5,000 comparison, p = 0.974 for the 1,000 and 10,000 comparison, and p = 0.970 for the 5,000 and 10,000 comparison). Based on these results, the combination of 1,000 cells per well and 1% matrix was selected because it permitted the use of a limited number of cells.

For the ovarian (SK-OV-3) model, the metabolic activity of the more proliferative cell conditions (1,000 cells per well with matrix) did not significantly differ, with values of 11.5 ± 0.4 UF for the 1% matrix, 11.5 ± 0.6 UF for the 2% matrix, and 11.1 ± 0.7 UF for the 3% matrix (p = 0.932 for the 1% and 2% comparison, p = 0.831 for the 1% and 3% comparison, and p = 0.539 for the 2% and 3% comparison) (Fig. 4, d). In conclusion, these three combinations have the potential to be used for drug efficacy evaluation.

Then, for the non-tumorigenic breast (MCF-10A) model, the more viable conditions (5,000 and 10,000 cells per well) presented a similar metabolic activity (p = 0.107) (Fig. 4, e). Considering these observations, selecting a concentration of 5,000 cells per well allowed the use of fewer cells per manipulation. There was no significant difference in metabolic activity among the three remaining conditions: 15.0 ± 0.1 UF for 1% matrix, 14.3 ± 1.8 UF for the 2% matrix, and 13.5 ± 0.6 UF for the 3% matrix (p = 0.939 for the 1% and 2% comparison, p = 0.180 for the 1% and 3% comparison, and p = 0.656 for the 2% and 3% comparison). As a result, these three conditions could be suitable for therapeutic drug screening.

Fig. 3
figure 3

Viability/Mortality profiles of 3D models with Live/Dead® test imaging at the end of the culture (day 14). (a) A-375 melanoma, (b) NCI-H1417 small cell lung cancer, (c) NCI-H1975 non-small cell lung cancer, (d) SK-OV-3 ovarian cancer, (e) MCF-10A non-tumorigenic breast and (f) VERO 76 kidney healthy cells were seeded at four cell concentrations, from 200 to 10,000 cells per well, with four matrix conditions, from 0% to 3%, in culture medium. To evaluate their viability, a Live/Dead® test was performed at day 14 of culture. Green fluorescence corresponds to viable cells and red fluorescence corresponds to dead cells. Acquisitions were made using Cytation10MV cell analyzer equipped with GFP and IP fluorescence cubes (Agilent, M = 4X, scale bar = 500 μm).

Full size image

Finally, with the kidney (VERO 76) model, the metabolic activity of the more proliferative cell conditions (1,000 cells per well with matrix) did not differ significantly, with values of 14.3 ± 0.8 UF for the 1% matrix, 14.3 ± 1.8 UF for the 2% matrix, and 15.1 ± 1.5 UF for the 3% matrix (p = 0.934 for the 1% and 2% comparison, p = 0.662 for the 1% and 3% comparison, and p = 0.759 for the 2% and 3% comparison) (Fig. 4, f). Therefore, these three conditions could potentially be used for drug screening.

Fig. 4
figure 4

Metabolic activity quantified using resazurin test at the end of the culture (day 14). (a) A-375 melanoma, (b) NCI-H1417 small cell lung cancer, (c) NCI-H1975 non-small cell lung cancer, (d) SK-OV-3 ovarian cancer, (e) MCF-10-A non-tumorigenic breast and (f) VERO 76 kidney healthy cells seeded at four cell concentrations, from 200 to 10,000 cells per well, with four matrix conditions, from 0% to 3%, in culture medium. To assess their metabolic activity, a resazurin-based test was performed at day 14 of culture. Resazurin is reduced into resorufin by metabolic active cells. Fluorimetry of resorufin was measured at λ=593 nm with fluorimetry module of Cytation10MV cell analyzer (Agilent). Data represent means ± SD. UF = “Units of Fluorescence”. A Kruskal-Wallis test has been performed.

Full size image

Spheroid conformation by scanning electron microscopy

Scanning Electron Microscopy (SEM) was used to assess the outside structure of the spheroids produced with the most adapted cell and matrix combination for each model.

All the selected cell and matrix concentrations led to the formation cohesive spheroids (Fig. 5). In more details, the melanoma (A-375) model had an ovoid shape and numerous free cells, which could escape from the spheroid, were detected on the outside of the micro-tissue (Fig. 5, a). Concerning the pulmonary models, a lot of free external cells could also be observed for the small cell lung cancer (NCI-H1417) subtype (Fig. 5, b), whereas the non-small cell lung cancer (NCI-H1975) subtype was completely encapsulated in matrix and perfectly round (Fig. 5, c). For the ovarian (SK-OV-3) model, matrix completely covered the cells composing the spheroid (Fig. 5, d). For the non-tumorigenic breast (MCF-10A) model, the outside of the spheroids was covered with very tightly stacked cells with a less visible matrix (Fig. 5, e). Finally, the kidney (VERO 76) micro-tissues were fully encapsulated in matrix (Fig. 5, f).

Fig. 5
figure 5

The 3D morphology of spheroids. Cells were seeded at optimal selected conditions of cell and matrix concentrations, and SEM images were taken with Jeol 6060 LV scanning electron microscope. A-375 melanoma spheroids (200 cells per well/2% matrix): (a1) M = 170X, (a2) M = 650X, NCI-H1417 small cell lung cancer spheroids (1,000 cells per well/1% matrix): (b1) M = 150X, (b2) M = 350X, NCI-H1975 non-small cell lung cancer spheroids (1,000 cells per well/2% matrix): (c1) M = 350X, (c2) M = 1 100X, SK-OV-3 ovarian cancer spheroids (1,000 cells per well/2% matrix): (d1) M = 350X, (d2) M = 1 500X, MCF-10A non-tumorigenic breast spheroids (5,000 cells per well/1% matrix): (e1) M = 180X, (e2) M = 370X and VERO 76 healthy kidney spheroids (1,000 cells per well/1% matrix): (f1) M = 190X, (f2) M = 500X.

Full size image

Anti-PARP drugs evaluation on cancer and non-tumorigenic models

To assess the relevance of our 3D cell culture panel for drug screening, we analyzed the sensitivity of our models to anti-cancer drugs. For these experiments, three clinically approved anti-PARP inhibitors, i.e. Olaparib, Rucaparib and Niraparib have been selected because they are targeted therapeutics that inhibit DNA repair pathways. To analyze anti-PARP efficacy, we selected three relevant 3D models: (i) SUM1315 TNBC cell line, which presents a BRCA1 mutation and was previously studied by our group, (ii) SK-OV-3 ovarian cancer cell line, which does not present a BRCA1 mutation, and (iii) MCF-10A non-tumorigenic breast cell line. The effects of the drugs on the three micro-tissues were analyzed after a 10-day treatment period at three in vitro effective anti-PARPs concentrations of 1 µM, 10 µM, and 50 µM.

Fig. 6
figure 6

The 3D cell culture models sensitivity to anti-PARP cancer treatments. Cells were seeded at optimal selected conditions of cell and matrix concentrations (SUM1315: 1,000 cells per well/2% matrix, SK-OV-3: 1,000 cells per well/2% matrix and MCF-10A: 5,000 cells per well/1% matrix). (1) Spheroid size and (2) Metabolic activity were measured after 10 days of exposure to three anti-PARP treatments: (a) Olaparib, (b) Rucaparib and (c) Niraparib, at three concentrations of 1 µM, 10 µM and 50 µM. Data represent means ± SD, where p < 0.05 (*) p < 0.01 (**), p < 0.001 (***); p < 0.0001 (****), according to Kruskal-Wallis test.

Full size image

First, Olaparib significantly impacted the spheroid size ratio of the two cancer models, compared to the control condition (Fig. 6. a1). SUM1315 breast cancer spheroids had size ratio values of 79.6 ± 4.3% (p = 0.0272) and 67.7 ± 3.0% (p < 0.0001), for 10 µM and 50 µM treatments, respectively. For SK-OV-3 ovarian cancer micro-tissues, a size ratio value of 74.9 ± 9.9% (p = 0.0224) was registered for 50 µM treatment. Olaparib did not affect the non-tumorigenic cell line at these tested concentrations. Regarding the metabolic activity rate analysis, a decrease to 45.3 ± 6.7% (p = 0.0006) was detected for the SUM1315 mutated model with 50 µM of Olaparib. In contrast, no significant toxicity was quantified at these concentrations on the SK-OV-3 non-mutated model. A decrease in metabolic activity was detected for the MCF-10A non-tumorigenic model (71.2 ± 7.8%, p < 0.0001) at 50 µM, revealing a slight impact of Olaparib (Fig. 6. a2).

Rucaparib treatment significantly decreased the spheroid size ratio in the SUM1315 model to 70.5 ± 5.0% (p < 0.0001) and 72.6 ± 2.9% (p = 0.0002) at 10 and 50 µM, respectively (Fig. 6. b1). For the other models, a proliferation rate of 61.0 ± 5.3% (p = 0.0010) was detected for the SK-OV-3 ovarian cancer model. Regarding toxicity, the cell metabolic activity decreased to 10.4 ± 0.1% (p = 0.0003) and 19.9 ± 15.0% (p = 0.0207) for the breast and ovarian cancer models, respectively, at the 50 µM dose, revealing an important impact of Rucaparib. In contrast, a 82.5 ± 7.0% (p = 0.0041) metabolic activity rate was detected in the MCF10A model at this concentration, indicating a minor impact (Fig. 6. b2).

Finally, for Niraparib, the spheroid size ratio decreased to 76.4 ± 4.3% (p = 0.0040), 78.3 ± 13% (p = 0.0017) and 79.9 ± 2.9% (p = 0.0330), from the lower to the higher concentration on SUM1315 mutated cancer model, respectively (Fig. 6. c1). For the SK-OV-3 non-mutated cancer spheroids, the spheroid size ratio decreased significantly to 58.6 ± 3.5% (p = 0.0082) and 51.5 ± 4.3% (p < 0.0001) at 10 and 50 µM, respectively. For the MCF-10A non-tumorigenic breast tissue, only the highest dose significantly impacted the spheroid size ratio, reducing it to 58.7 ± 7.1% (p = 0.0176). Metabolic activity was considerably reduced for SUM1315 spheroids to 44.7 ± 5.1% (p = 0.0336) and 10.0 ± 0.1% (p = 0.0001) at 10 and 50 µM, respectively (Fig. 6. c2). An important reduction in metabolic activity to 7.3 ± 0.3% (p = 0.024) was observed for the 50 µM concentration in the SK-OV-3 ovarian cancer model. For the MCF-10A non-tumorigenic breast tissue, the viability reduction was relatively minor, with values of 81.8 ± 8.4% (p = 0.0227) and 69.1 ± 15.3% (p = 0.0010), for the two highest doses, respectively.

Discussion

The development of pharmaceutical drugs requires innovative advancements at the cellular stage to improve predictability and, consequently, reduce the use of animals in screening assays. In this context, 3D cell culture models are particularly relevant as they could permit to rapidly obtain results regarding drug penetration, growth impact and toxicity on micro-tissues17. Indeed, they are highly similar to tissues in terms of conformation, cell-cell and cell-matrix interactions, as well as genetic expression profile7. Consequently, this approach enables to obtain more predictive data concerning the concentration of drug causing cytotoxic and cytostatic effects on multi-layer tissues. Finally, the use of 3D cell culture could lead to a significant reduction in animal testing, thus allowing a more ethical pharmaceutical development process.

In this context, the focus of these studies was the development of a simple, cost-effective method adapted from the liquid overlay technique14 for generating spheroids that could be used in any standard cell culture laboratory without requiring specialized equipment or advanced technical expertise. It only requires determining the concentrations of cells and matrix, which allows the production of single, size-controlled, and viable spheroids for up to 14 days of culture. The concentrations of cells and matrix selected for the six new models were: 200 cells per well with 2% matrix for A-375 melanoma model, 1,000 cells per well with 1–3% matrix for NCI-H1417 small cell lung cancer model, 1,000 cells per well with 1% matrix for NCI-H1975 non-small cell lung cancer model, 1,000 cells per well with 1–3% matrix for SK-OV-3 ovarian cancer model, 5,000 cells per well with 1–3% matrix for MCF-10A non-tumorigenic breast model and 1,000 cells per well with 1–3% matrix for VERO 76 healthy kidney model. Our panel was designed to complement our previous breast cancer models (MDA-MB-231 and SUM1315) and combine non-tumoral and tumor-derived 3D models from various tissues. This provides a broader range of biological contexts and enables the evaluation of drug responses across different genetic backgrounds and tissue origins. As demonstrated, this protocol is highly adaptable and could be modified to be compatible with all types of cells, including healthy and cancerous cells, as well as those derived from human and other animal tissues.

For all these models, a SEM structure characterization has been performed. For melanoma spheroids, a quite ovoid shape has been reported, which could be related to the characteristic asymmetric shape of this cancer18. Moreover, the numerous free cells detected on the outside of the micro-tissue could reflected a tumor aggressive state, which clearly correlates with the poor prognosis of this skin cancer subtype19. For lung cancer models, small cell one exhibited a lot of free external cells on the outer layer, which could reflect an aggressive subtype, whereas non-small cell one was encapsulated in matrix. This phenotype is characteristic of less invasive cancers, associated with better prognosis. These results seem to be in adequation with the clinical data available for lung cancer20, which classifies small cell lung cancer as a more lethal one. Subsequently, for ovarian model, no distinct cells were detected on the surface of the spheroids, and they were quite round, which is a characteristic of non-long distance metastatic tumor. Indeed, ovarian cancers typically spread to nearby tissues, such as uterus or fallopian tubes21. For the non-tumorigenic mammary model, the surface was covered with very tightly stacked cells and matrix similar to the glandular tissue from which the cells were originated22. For the kidney model, the micro-tissue was fully encapsulated in matrix, which confirmed the non-pathologic state of the model. All the SEM analysis seem to be correlated with the available clinical data. It appears that the potential more aggressive models, such as the melanoma and the small cell pulmonary spheroids present on their surface more cells, which are not embedded in the matrix, corresponding to potential metastatic cells. Moreover, the less aggressive and the healthy models present matrix to their surface representing poor or non-metastatic tissues.

Subsequently, the applicability of drug screening on these models has been carried out on three selected micro-tissues with three different anti-PARP treatments. These drugs are targeted therapeutics that inhibit DNA repair pathways. Indeed, in healthy tissues, DNA breaks require PARP enzyme and BRCA1 protein to be repaired23. In cancer cells, their high proliferation rate results in a large number of DNA breaks that cannot be repaired as rapidly as usual. Therefore with anti-PARP treatment, this phenomenon leads to their accumulation. Furthermore, according to the synthetic lethality principle, in the case of a BRCA1 mutation affecting one of the DNA repair pathway, an excess accumulation of DNA breaks occurs with anti-PARPs, resulting in a significant degree of toxicity and subsequent cell death28,29,30. Consequently, to study the sensitivity of our models, we have selected a breast BRCA1 mutated cancer (SUM1315) model14, an ovarian BRCA1 non-mutated cancer (SK-OV-3) model and a non-tumorigenic breast (MCF-10A) model. First, this study demonstrates that our spheroids respond to treatments, with different degrees of response depending on the molecules tested. Regarding the spheroid size ratio, Olaparib had a stronger overall impact on SUM1315 breast mutated cancer model than on SK-OV-3 ovarian non-mutated cancer model (p = 0.001) and MCF-10A non-tumorigenic breast model (p < 0.001). Similarly, Rucaparib had a more significant cytotoxic effect on the diameter ratio of SUM1315 spheroids than on SK-OV-3 (p = 0.074) and MCF-10A (p = 0.024) models. For these two treatments, there were no significant differences in micro-tissue size between SK-OV-3 and MCF-10A models (p = 0.979 and p = 0.287, respectively). Niraparib treatment equally impacted the size of all models (p = 0.878 for SUM1315/SK-OV-3 comparison, p = 0.234 for SUM1315/MCF-10A comparison and p = 0.500 for SK-OV-3/MCF-10A comparison). Regarding metabolic activity, Olaparib has significantly impacted the SUM1315 BRCA1-mutated cancer model more than SK-OV-3 non-mutated cancer model (p = 0.004). Whereas Rucaparib and Niraparib led to similar decreases in metabolic activity among the two cancer models (p = 0.078 and p = 0.619, respectively). As expected, anti-PARPs exhibited a greater impact on the BRCA-mutated cancer model than on the non-mutated model, while non-tumorigenic breast spheroids were largely unaffected by the treatment. Additionally, this efficacy study confirms the correlation of the obtained results with the available clinical data. Indeed, Olaparib, Rucaparib and Niraparib have a clinical approval for ovarian cancers24,27, and a large panel of clinical trials are undergoing for breast ones25. To date, Olaparib and Talazoparib (not included in the study) have received approval from the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) for HER-2 negative breast cancer subtype25,27. Consequently, this presented cell culture panel suggests that Rucaparib and Niraparib should have a strong cytotoxic effect on mutated breast cancer cells while remaining inactive on healthy cells, thereby highlighting their clinical relevance.

In conclusion, our protocol for producing 3D cell models results in the formation of size-controlled, viable and active spheroids that are highly reproducible. Furthermore, the technique is compatible with automated analysis, consequently making it suitable for Medium Throughput Screening (MTS). These models are promising alternative to animal testing in terms of predicting the drug’s cytostatic and cytotoxic impacts. This easily adaptable method could be applied to all types of tissue, regardless of the origin or the pathologic state. Moreover, it is possible to determine the efficacy and the toxicity of a large number of therapeutic molecules in a single assay, as well as investing the combination and duration of treatments. Indeed, monoculture models are used to obtain the more homogeneous and reproducible micro-tissues. However, they do not allow the assessment of the impact of all drugs, such as immunotherapies. To address this limitation and further complete our panel, additional cell types, i.e. fibroblasts31,32, immune cells33,34 and endothelial cells35,36 could be integrated into a co-culture to mimic the microenvironment more closely. These additions have the potential to expand the applicability and the relevance of these micro-tissues to other preclinical studies.

Methods

Cell lines

Cell lines were obtained from ATCC (Virginia, USA) and absence of mycoplasma contamination check (Mycoplasmacheck test, Eurofins Genomics, Luxembourg, Luxembourg) was performed. Human cell lines were authenticated by Short Tandem Repeats (STR) profiling (Cell Line Authentication Service, Eurofins Genomics, Luxembourg, Luxembourg).

Cell culture

SK-OV-3 (HTB-77) and NCI-H1975 (CRL-5908) cells were cultured in RPMI 1640 medium (Gibco, Dublin, Ireland) supplemented with 10% decomplemented Fetal Calf Serum (Eurobio Scientific, Paris, France) and 20 µg/mL Gentamycin (Panpharma, Paris, France). NCI-H1417 (CRL-5869) cells were cultured in RPMI 1640 medium (Gibco, Dublin, Ireland) supplemented with 10% decomplemented Fetal Calf Serum (Eurobio Scientific, Paris, France) and 1X Penicillin Streptomycin (Gibco, Dublin, Ireland). A-375 (CRL-1619) cells were cultured in DMEM medium (Gibco, Dublin, Ireland) supplemented with 10% decomplemented Fetal Calf Serum (Eurobio Scientific, Paris, France) and 20 µg/mL Gentamycin (Panpharma, Paris, France). VERO 76 (CRL-1587) cells were cultivated in DMEM medium (Gibco, Dublin, Ireland) supplemented with 10% decomplemented Fetal Calf Serum (Eurobio Scientific, Paris, France) and 1X Penicillin Streptomycin (Gibco, Dublin, Ireland). MCF-10A (CRL-10317) cells were cultured in DMEM/F-12 medium (Gibco, Dublin, Ireland) supplemented with 5% decomplemented Horse Serum (Gibco, Dublin, Ireland), 20 ng/mL of Epidermal Growth Factor (Sigma Aldrich, Saint-Louis, Missouri, USA), 0,5 µg/mL Hydrocortisone (Sigma Aldrich, Saint-Louis, Missouri, USA), 10 µg/mL Insulin (Gibco, Dublin, Ireland), 100 ng/mL Cholera Toxin (Sigma Aldrich, Saint-Louis, Missouri, USA) and 1X Penicillin Streptomycin (Gibco, Dublin, Ireland). Cell lines were incubated at 37 °C under 5% CO2 in a humid incubator. To minimize genetic mutation and preserve experimental reproducibility, the cells were used within the generally accepted range of passages below 20.

3D cell culture models production by liquid overlay technique

3D cell culture protocol was processed according to the method described by Dubois et al.14. Consequently, Geltrex matrix was used because it had already been validated in this previous study. Indeed, this matrix has an appropriate composition in ECM proteins, including laminin, collagen IV, entactin, and heparin sulfate proteoglycans. Additionally, there is a low variation in protein concentration among the different batches (approximately 15 mg/mL). For the production protocol, cells were seeded in an “ultra-low-attachment” (ULA) round-bottom 96-well microplate (catalog no. 4520, Corning, NY, USA), and agitated on a microplate shaker at 200 rpm for 20 min. After 24 h of incubation, a cold (4 °C) solution of 100% Geltrex Lactic-Dehydrogenase-virus-free (catalog no. A1413202, Gibco, Dublin, Ireland), diluted in the appropriate culture medium, was dispensed on the aggregated cells. The plate was then agitated a second time as previously described. After an overnight incubation, the spheroids were formed and ready for the experiment.

Growth evaluation

Spheroids were imaged using bright field microscopy module of Cytation10MV cell analyzer (Agilent, Santa Clara, California, USA, M = 4) paired with automatic incubator BIOSPA8 (Agilent, Santa Clara, California, USA). “Cellular analysis” tool of Gen 5.03 software (Agilent, Santa Clara, California, USA) was applied to a stitched picture to calculate the object size of each well (background = light, min. object size = 100 μm, max. object size = 1,200 μm). Spheroid size was recorded for each condition at the 1st (day of effective formation), 3rd or 4th or 5th, 7th, 9th or 10th and 14th day of culture (D1, D3/4/5, D7, D9/10 and D14) allowing to set growth kinetic curves.

Viability assessment

The spheroid cell viability was monitored at day 14, the end of the experiment, using Live/Dead® fluorescent microscopy kit (Invitrogen, Waltham, MA, USA). Spheroids were transferred in an 8 well µ-Slide (Ibidi, Gräfelfing, Germany) to be incubated for 45 min at room temperature, protected from light, in a working solution containing 1 µM of calcein-AM (green fluorescence) and 2 µM of ethidium homodimer-1 (ethD1, red fluorescence) in Phosphate Buffer Saline (PBS). They were imaged with the fluorescent microscopy module of the Cytation10MV cell analyzer (Agilent, Santa Clara, California, USA, M = 4, fluorescent filters = Green Fluorescent Protein and Propidium Iodide).

Metabolic activity assay

Additionally, spheroid metabolic activity was measured at day 14, by transferring each spheroid in 100 µL of a sterile 60 µM resazurin solution in PBS. After 16 h incubation at 37 °C, the fluorescence intensity of resorufin (593 nm) was measured using fluorimetry module of Cytation10MV cell analyzer (Agilent, Santa Clara, California, USA, gain = 50 or 70). Blank normalization was applied on data obtained from spheroid free wells.

Structure analysis

For Scanning Electron Microscopy observations, spheroids were fixed 1 h with 1.6% glutaraldehyde and ruthenium red 0.15% in 0.2 M sodium cacodylate buffer, pH 7.4. They were then washed 30 min in sodium cacodylate buffer (0.2 mol/L, pH 7.4) and post-fixed 1 h with 1% osmium tetroxide in same buffer. After rinsing for 20 min in distilled water, dehydration by graded ethanol was performed from 25° to 100° (10 min each) to finish in hexamethyldisilazane (HMDS) for 10 min. Samples were mounted on stubs using adhesive carbon tabs and coated with chrome (Quorum Q150 TES, UK). Observations were carried out using a Scanning Electron Microscope 6060-LV (JEOL, Tokyo, Japan) at 5 kV with a secondary electron detector.

Anti-PARP treatment solubilization, exposure and analysis

Olaparib (catalog no. FO331222, Carbosynth, Compton, UK), Rucaparib (catalog no. B2693-104639, BocSciences, Shirley, NY, USA) and Niraparib (catalog no. B0084-408479, BocSciences, Shirley, NY, USA) were solubilized in DMSO to prepare 100 mM stock solution. Dilutions were prepared in adapted medium for final treatment concentrations of 1, 10 or 50 µM. The final DMSO concentration remained constant at 0.1% in all tested conditions (including Control). Spheroids were treated at day 3 for 10 days. Percentage of proliferation is the size ratio between day 13 and day 1, compared to control condition without treatment for each model (= 100% spheroid size ratio). Percentage of metabolic activity is the blank normalized value, compared to control condition without treatment for each model (= 100% metabolic activity).

Statistical analysis

All experiments were repeated three times, with six spheroids per experiment. Results were expressed as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism and Jamovi. Differences were assessed using the Kruskal–Wallis test, followed by Dunn’s post hoc test with multiplicity-adjusted p-values for DCSF pairwise comparisons. Results were considered statistically different when p < 0.05 (*). Stronger differences were noted as follow: p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****).