A polysaccharide-based hydrogel platform for tumor spheroid production and anticancer drug screening

Abstract

Extracellular matrix mimics are still needed to grow cancer cells in 3D environments and study their evolution in vitro while precisely controlling relevant features. Most models currently use collagen, which is biomimetic but degrades quickly, or artificial polymers, which can be chemically modified but remain stiff. Herein we introduced a soft, non-adhesive, and resistant hydrogel platform for tumor spheroid production using a polysaccharide-based formulation. To ensure micro-structuring of the hydrogel and enable spheroid formation, 3D printed molds consisting of a network of 200-µm-diameter micropillars were used to generate microstructured hydrogel constructs that fit into a multi-well plate. This platform was validated for drug testing using three cancer cell lines (A673, MCF7 and U87) and 2 anticancer drugs (doxorubicin and paclitaxel). Drug response was assessed through bright-field microscopy monitoring and viability measurements after 48 h of treatment. This study validates the use of pullulan-dextran hydrogels for spheroid formation, combined with in situ drug screening.

Introduction

Cancer progression remains a significant threat, and precisely characterizing the influence of each component of the tumor microenvironment (TME) on its development continues to be a complex challenge. The TME is composed of a extracellular matrix (ECM) enriched in collagen I, laminin, and fibronectin, of supportive cells such as cancer-associated fibroblasts (CAFs) or mesenchymal stromal cells (MSCs), of proteins that are either secreted by surrounding cells or contained in the matrix, and of a vascular compartment. Current cancer models aim to reproduce the complexity of the TME while being able to differentiate its various features. To this end, three-dimensional (3D) cell culture has been introduced in the last years to recapitulate the organization of tumor cells without the need for animal models. These models can serve as platforms to study the interplays between cancer cells and specific characteristics of the TME by independently selecting the different components to integrate^1,2.

The choice of the matrix is driven by its mechanical properties, ease of manipulation, and the geometry that it provides for cancer cells. The material must fall within the physiological range of stiffness that cells experience in vivo, to trigger the same mechanobiological behaviors^3,4. Additionally, the mechanical properties will influence the choice of production technique, presenting additional constraints when using methods like bioprinting, which demand significant expertise and expensive equipment.

Most cancer models agree on the importance of accurately recapitulating the 3D cellular organization, with spheroids becoming the gold standard. Spheroids can be produced using different techniques such as hanging drop^5,6, magnetic levitation⁷, or ultra-low attachment wells^8–12, before being transferred to the setup of interest, mostly constituted of collagen^5–7or fibrin^8–10. These materials effectively mimic the ECM, allow for precise control of their mechanical properties, and are well-known for their biocompatibility¹³. However, they are quickly degraded by cells, especially when supportive cells are added to the model, as this fosters the cancer cells aggressiveness directly linked to the active remodelling of the matrix^14–16. Consequently, while these materials are tailored for such analyses, they require an additional step of transfer after spheroid formation and prevent long-term culture because of their rapid degradation that also impairs the observation of samples with time.

Therefore, systems that allow for both the production and study of spheroids in situ represent an interesting alternative with a more straightforward process. To this extent, droplet microfluidic techniques allow performing large-scale drug screenings on single spheroids^17,18, while utilizing printing or molding of ultra-low attachment supports from antiadhesive materials such as PDMS^6,19–21or agarose^22–24. Finally, several studies leverage the inherent porosity of hydrogels to grow spheroids in a biomimetic environment^{2,11,12,25,26}.

In this study, we present a biomimetic 3D hydrogel model that uses a combination of two polysaccharides, pullulan and dextran. These biomimetic materials are produced by the fungus Aureobasidium pullulan and lactic acid bacteria such as Leoconostoc mesenteroides respectively. Polysaccharide-based hydrogels are already commercialized for in vitro applications using simple molding or 3D-printing and show great biocompatibility and tunability^27–29. More particularly, pullulan-dextran (PuD) hydrogels have been developed and were evaluated both in vitro^30–32and in vivo^33–35. They are cost-effective and combine relevant mechanical properties, resistance to cell degradation, with ease of production and imaging. Indeed, they are transparent, which makes them compatible with microscopy monitoring, and in the case of thick constructs loaded with high cell densities, they can undergo clearing techniques³². Consequently, sample staining can be performed in situ, facilitating characterization and follow up of the model. Live cell analyses, such as viability and protein secretion assays can also be conducted within these hydrogels due to their permeability that ensures protein diffusion through the matrix.

Herein, PuD hydrogels are proposed as non-adhesive supports to form tumor spheroids in a TME mimic. For that, an initial fabrication step using 3D printed molds to pre-form microwells during the gelation of PuD hydrogels was implemented. Spheroids from Ewing sarcoma (A673), glioblastoma (U87) and breast cancer (MCF7) were then formed and treated in situ with two anticancer drugs, namely doxorubicin (DOX) and paclitaxel (PTX). DOX and PTX have been reported to be efficient on various cancer and organs, although they act differently. DOX can intercalate within DNA strands, which leads to instabilities and further elimination by the repair system^36,37. On the other hand, the main mechanism of action of PTX is mediated by microtubule stabilization, impairing actin remodelling and mitosis^38,39. Both drugs have also been shown to occasionally induce apoptosis by alternative pathways⁴⁰, or to promote the formation of reactive oxygen species^36,41.

In this work, we introduce PuD hydrogels to form a new biomimetic platform as a tool for studies aimed at drug testing with 3D tumor models. This hydrogel is versatile, biocompatible, highly stable, facilitating long-term studies, and adaptable to most imaging techniques thanks to its transparency. By using this model to grow three different cancer cell lines, we assessed their response to two different drugs in 3D conditions, which validates the model as a platform for spheroid growth and anticancer drug testing.

Results

Platform design: shaping the polysaccharide hydrogel with microwells

The hydrogel platform for spheroid production was developed to achieve cost-effective and rapid synthesis, providing a microenvironment to cancer cells and enabling their growth in 3D as spheroids (Fig. 1A). Using an already established formulation^31,32,42, hydrogels were produced in sterile conditions to obtain batches of PuD hydrogels of ca. 1 cm diameter and 1.5 cm height (Fig. 1B), storable at 4 °C for up to 3 months, and transparent for imaging. They were made fluorescent by incorporating 0.05% w/w FITC-dextran in their composition for morphological analysis using confocal microscopy. To better account for the tridimensional aspect of the tumor, we aimed at producing spheroids. Using PuD as a non-adhesive material, we created the internal architecture of the hydrogels using a 3D printed mold with micropillars. This method of production is simple and allows for easy design modifications. Moreover, PuD crosslinks in 20 min at 50 °C without need of any toxic solvent and can be easily unmolded with a custom-made puncher and ring setup (Fig. 1B). As a result, each hydrogel contained 37 microwells of 287 ± 15 µm in diameter and ca. 200 µm high, that were disposed in three concentric circles (Fig. 1C&D). An analysis of the reconstituted profiles of microwells showed their funnel shape with measured dimensions. Microwells diameters are normally distributed, with a narrow gaussian peak for well size as well as circularity. A final step of washing following crosslinking allowed to release the entrapped ions and to control the swelling of the hydrogel and therefore, the final size of the pattern. Then, the hydrogels were UV-C treated (254 nm for 1h) to be usable for cell culture over long-term periods.

Fig. 1.

Development of the polysaccharide platform. (A) Schematic of the synthesis. Pullulan and dextran were dissolved in water, and a solution of sodium trimetaphosphate (STMP) was added to trigger a chemical crosslinking supported by heating. The hydrogels were then suitable for cell culture and adapted to 48 well plates. Created with Biorender. (B) Pictures of the transparent hydrogels produced with this process, using a 3D-printed mold (in black, with scale) and a puncher (in black, bottom image and side view) with removable rings to facilitate unmolding (transparent, top image). (C) Hydrogels doped with FITC-dextran were imaged by confocal microscopy to evidence the robustness of the pattern, as well as the dimensions of the microwells profile, that are depicted on a scheme to ease comprehension (n = 3 for 3 different batches). Scale bar 500 µm. (D) A radius analysis on over 150 microwells showed an average well radius of 143 ± 7 µm with an aspect ratio under 1.1.

Spheroid formation in polysaccharide microwells

The synthesis described above enabled the production of hydrogels that can be contained within a 48-well plate. These hydrogels had a U-shaped design with 37 microwells at the bottom of their container (Supplementary Fig. S1). This container allowed for the dropwise addition of cell solutions, maximizing the number of cells that fell randomly across all microwells. Besides, the funnel shape of the microwells ensured that no cells were trapped between them. After cell deposition, a centrifugation step was performed to help cells fall into microwells where they aggregated and further formed cohesive spheroids in less than 3 days (Fig. 2A). A673, MCF7 and U87 cell lines were used for this study. Spheroids were analyzed by bright field imaging using a plate reader to automatize acquisition and their shape and size were monitored over the culture period (Fig. 2B). Circularity and aspect ratio were selected as complementary parameters to quantify the roundness of spheroids: circularity is a ratio between the area and the perimeter, making it sensitive to structures with irregular borders, such as loosely formed spheroids for example. Aspect ratio is the ratio between the largest and the smallest diameter, allowing to detect isotropic objects deformation. Both parameters reach 1 for a perfectly round object. Three different trends could be observed (Fig. 2C): A673 formed cohesive spheroids in 3 days, which grew rapidly until filling the whole well size at day 5, giving a quite uniform population of 250-µm-diameter circular spheroids. On the other hand, MCF7 formed rather loose spheroids with no distinct circular shape, often aggregating or growing in a clumsy manner. This resulted in a decrease in their spherical aspect between day 3 and 7. Besides, we could observe the formation of multiple yet smaller spheroids per microwell for MCF7 cells. However, significant variations were noted, including very small cellular aggregates as well as “mega-spheroids” formed by the aggregation of several spheroids. U87 demonstrated a more typical growth pattern, with spheroids increasing in size throughout the maturation process and becoming more spherical over time. They also tended to grow slower than A673 cells and occasionally formed multiple small spheroids in a single microwell. In addition, all spheroids showed a darkening over time, indicating an increase in cell density within 4 days. Based on these findings, we opted to seed A673 cells at a density of 30.000 cells per hydrogel, while doubling this quantity for MCF7 and U87 cells.

Fig. 2.

Tumor spheroid growth. (A) Bright-field monitoring of the spheroids with time. Ewing sarcoma (A673), breast adenocarcinoma (MCF7) and glioblastoma (U87) cell lines formed spheroids in 3–5 days. Scale bar 500 µm (B) Optimisation of the spheroids size with the initial density. Image analyses of the spheroids for one week showed a significant increase with both time and cell density. Data presented as average ± standard deviation and compared using ordinary two-way ANOVA (alpha = 0.05) with Tukey’s post-hoc multiple comparisons versus the “15k D3” condition for each cell type (ns, not significant, *, p-value < 0.05, **, p-value < 0.01, ***, p-value < 0.001, ****, p-value < 0.0001). n > 56 spheroids from different wells of the same seeding for all groups. (C) Evolution of the spheroids shape with time. Spheroids shape was evaluated using circularity and aspect ratio, which underlined three different trends for the different cell types (as shown by arrows). Each point represents a cell density and is the average of over 56 values. The image with error bars is shown in Supplementary Fig. S2 to show variability.

Proof-of-concept of spheroids use with an anticancer treatment

Once spheroids were formed, drug response could be assessed. As an initial step, A673 spheroids were treated with the anticancer drug DOX. Spheroids were grown for 3 days before being treated with drug diluted in culture medium for 48 h. Additionally, spheroids were imaged to analyze their growth under treatment (Fig. 3A). Like observations without treatment, lowest doses exhibited significant growth within 2 days, whereas intermediate doses (e.g. 0.02 and 0.39 µM) showed reduced growth. The highest doses were lethal for A673 spheroids, leading to a loss of cell cohesion and a collapse of spheroids with cells spreading throughout each microwell. An analysis of spheroid size with increasing dose was further performed, confirming a distinct transition between spheroids that continued to grow up to twice their initial size (100% growth, such as control and lowest doses of drug) and spheroids that exhibited minimal growth over 2 days (0%). The highest dose of 100 µM consistently resulted in the collapse of spheroids, making them appear as a hoop of dead cells with a darkening of the covered area. The presence of such hoops was used as a criterion to exclude data points from the analysis. It was then possible to adjust the growth curve and determine a half maximal inhibitory concentration (IC50) of 0.03 µM (Fig. 3B). Finally, spheroids were recovered from the microwells after treatment, stained for F-actin, and imaged by confocal microscopy (Fig. 3C). DOX is naturally fluorescent and binds to DNA in the nuclei. Yet, because of the very broad excitation and emission spectra of DOX and Alexa Fluor 555, small interferences took place, that explains why the control showed a weak signal at the emission wavelength of DOX. For treated spheroids, we observed a decreasing signal for F-actin with increasing doses, accompanied by a progressive disorganisation of the filaments. Additionally, DOX-stained circular shapes distinctly for the two highest doses, whereas the stained area became less distinct for lower doses, with a dark central area likely indicating limited drug diffusion into the cohesive spheroid core. This was confirmed by the ability of inner cells to adhere to a substrate and spread after having received a DOX treatment (Supplementary Fig. S3).

Fig. 3.

Doxorubicin (DOX) treatment on A673 spheroids. (A) Bright-field monitoring of the spheroids size before and after 48 h of DOX or control treatment. Scale bar 500 µm (B) IC50 estimation using image analysis. The analysis of spheroids size evolution with the different doses showed a clear viability decrease, yet mitigated by the relapse at the highest doses that could be explained by the loss of cohesion of the spheroid when all cells were dead. The dark blue point was excluded for the fit based on its bright field appearance. Pooled data of two independent experiments (light grey, 3–6 wells per dose and experiment) are represented by the average value with its standard error (blue), along with a 4-parameter sigmoidal fit and its 95% confidence interval. (C) Confocal imaging of single spheroids at different DOX doses. Actin and DOX levels were observed after staining, highlighting a disturbed actin network for intermediate doses, and complete loss of cohesion for the most lethal doses. On the contrary, DOX-stained nuclei for highest doses and struggled to penetrate at the inner core of the best-preserved spheroids. The viability of preserved cells is shown in Supplementary Fig. S3. Scale bar 200 µm.

IC50 analyses by metabolic activity and viability

Then, drug testing was conducted across the three cancer cell types, A673, MCF7 and U87, using the two anticancer drugs, DOX and PTX. Drug response was assessed using the metabolic activity of the cells as an indicator of drug efficacy. Additionally, we compared the behavior of each cell type organized as spheroids (3D) with the same cells cultured in a monolayer (2D) (Table 1). For DOX treatment, all three cell types exhibited IC50 values between 0.05 and 0.6 µM. The values were similar in 2D and 3D cultures, but higher in 3D for MCF7 and U87, indicating increased resistance to treatment, while A676 cells showed higher resistance in 2D. This higher sensitivity in 3D was also observed with PTX for A673 and MCF7, with a twofold increase in IC50 between 2 and 3D. Besides, three distinct trends emerged from these IC50 curves (Fig. 4). All conditions treated with DOX appeared to exhibit a viability plateau higher than 0%, despite the spheroids loose appearance under bright field microscopy. These spheroids were still able to spread and grow afterwards (Supplementary Fig. S3A). An increase in metabolic activity also appeared in every cell line treated with PTX in 2D just before IC50, as well as for MCF7 and U87 treated with DOX at high doses, after the IC50 transition.

Table 1.

IC50 values for 2D and 3D conditions for 3 different cell types and 2 different drugs.

	DOXORUBICIN (DOX) by Alamar			PACLITAXEL (PTX) by Alamar			PACLITAXEL (PTX) by LDH
	A673	MCF7	U87	A673	MCF7	U87	A673	MCF7	U87
IC503D (µM) 95% CI	0.23 0.15–0.34	0.20 0.12–0.35	0.56 0.34–1.12	5.60 NA-7.77	12.7 6.48–44.2	14.4 12.2–21.4	3.12 NA-NA	20.6 NA-27.0	30.2 10.2-NA
IC502D (µM) 95% CI	0.46 0.35–0.61	0.05 0.03–0.08	0.39 NA-NA	12.4 9.73–15.6	22.3 19.6–25.1	13.9 NA-15.8	8.44 5.15–15.5	17.2 16.3–18.1	14.3 10.7–19.6
Ratio 3D/2D	0.50	4.00	1.43	0.45	0.57	1.04	0.37	1.20	2.11

All the experiments were normalized and pooled, and IC50 values with their 95% confidence interval (CI) were extracted from a top-locked 4-parameter sigmoidal fit. Data from at least 3 independent experiments were collected, with 3–6 values per dose and experiment. NA: not available.

Fig. 4.

Versatility of the platform for drug screening. (A) DOX treatment of three different cell types in 2D monolayers (blue squares) and in the polysaccharide-based 3D platform (orange circles) and analysed using a resazurin-based assay. (B) PTX treatment of three different cell types in 2D monolayers (blue squares) and in the polysaccharide-based 3D platform (orange circles) and analysed using a resazurin-based assay. The metabolic activity was assessed after 3 days of maturation and 2 days of treatment. Three to six independent data points per dose were acquired for each experiment and normalized against the top plateau. The data were then pooled across independent experiments with n = 6, 5, 6, 5, 5, 5 in 2D and n = 7, 3, 3, 2, 3, 4 in 3D for the conditions A673 DOX, MCF7 DOX, U87 DOX, A673 PTX, MCF7 PTX and U87 PTX respectively. (C) PTX treatment of three different cell types in 2D monolayers (green squares) and in the polysaccharide-based 3D platform (orange circles) and analysed using a lactate dehydrogenase-based assay. The viability was evaluated after 3 days of maturation and 2 days of treatment. Two to four data points per dose were acquired for each experiment and normalized against the top plateau. The data were then pooled for 3 independent experiments across all conditions. For each dose, the average value and its standard error are represented, along with a top-locked 4-parameter sigmoidal fit with its 95% confidence interval. The confidence intervals diverged for U87 DOX and for A673 PTX and are therefore missing.

Experiments conducted with LDH confirmed that IC50 values were close between 2D and 3D cultures for PTX treatment. Specifically, A673 showed a threefold decrease in IC50 between 2D and 3D, while U87 exhibited a twofold increase and MCF7 remained constant. These trends align with AlamarBlue data. No drop in viability was observed before IC50 in 2D conditions with LDH, confirming that it might be a metabolic increase before death.

Discussion

We have developed a hydrogel platform for controlled spheroid production using a non-adhesive natural polysaccharide-based biomaterial. To prepare the hydrogels, we combined a PuD-based hydrogel with tailor-made 3D printed molds. The gel solution covalently crosslinked in 20 min at 50 °C using a harmless crosslinker (STMP) and basic activation with sodium hydroxide, adhering to green chemistry principles with a very simple process. The gel solution was poured into the molds immediately after mixing with the crosslinker, and gentle pressure was applied using glass slides to ensure that the viscous liquid fully impregnated the molds. After gelation, 40 hydrogels each containing 37 microwells within a U-shaped structure were obtained. This production capacity was solely limited by the number of holes in the mold, allowing significant room for scaling up. The analysis of microwell dimensions showed that the pattern was well retained in the hydrogels, and the molds were easy to wash and reuse. The final hydrogel was transparent, and its swelling was controlled by adjusting the ionic strengths of the solutions used in the washing steps to ensure good pattern retention for 3D cell culture. Besides, it was non-adhesive, which helps form spheroids, and prior studies with a similar hydrogel composition exhibited relevant biomechanical properties to mimic TME with G’ values between 0.3 and 3 kPa³¹. Besides, porous PuD hydrogels were recently reported by some of us to induce hepatic cell assembly within the pores for up to 3 weeks without material degradation³⁵. The effect of hydrogel porosity on spheroid growth has been studied by Lemarié et al¹²., which compared different formulations of hydrogels as scaffolds to grow spheroids of induced pluripotent stem cells in close contact. They evidenced that porosity could promote spheroid growth by easing their spreading without need for matrix degradation in these porosities. Bonanini et al. showed that porous hydrogels such as collagen I were also privileged supports to study cancer-associated angiogenesis¹¹. In a commercial chip, they studied the influence of a hepatic cancer spheroid on the sprouting of a vascular bed in collagen. They proved the perfusability of their setup by imaging the progressive diffusion of fluorescent dextran throughout the spheroid thanks to the connectivity of their vascular bed with prevascularized spheroids. Such system is of great interest to study the evolution of a tumor spheroid in a more biomimetic environment that comprises a dynamic component, which can be used for drug screening perspectives for example. Using hydrogels, we produced spheroids from different tissues origin in a soft microenvironment⁴³compared to artificial polymers such as PDMS or agarose, while also preventing rapid degradation, making it an interesting compromise. Indeed, PDMS is very easy to template and prevents cell adhesion, which makes it an interesting material to build microwells for spheroid generation¹⁹, and to envision industrial scale-up perspectives¹². For example, An et al. presented an interesting way to engineer wells of different shapes by adding non-crosslinked gel solutions on liquid PDMS, and playing with the relative surface tensions of the different liquids²⁰. It allowed them to create wells with different curvatures, in which they investigated the response of cancer spheroids to drug screenings. By treating human kidney carcinoma cells with PTX, they also evidenced a loosening of spheroids with treatment, that they confirmed with a Live/Dead assay. Finally, PDMS is also widely used to create microfluidic chips, that can be patterned to integrate individual spheroid chambers to subject them to flow²¹. An interesting alternative is agarose, whose stiffness is more adapted to biological systems, and which prevents any unwanted therapeutic absorption. This hydrogel is antiadhesive and covers a wide range of mechanical properties depending on its formulation. Thanks to an easy crosslinking process at room temperature, it can be molded or patterned using stamps, which makes it a relevant choice to engineer microwells for spheroid production^22,23.

Patterning microwells in our polysaccharide-based platform allowed us to produce 250-µm-diameter spheroids in a consistent way. Interestingly, spheroid growth assessed by optical microscopy was different for all three cell types (Fig. 2). A673 spheroids grew faster than MCF7 and filled the whole well at D5, therefore not growing much more afterwards. U87 had a very classical growth pattern, increasing regularly over time and depending on initial seeding density. These varying spheroid morphologies across different cancer cell types have already been described before and are expected to underline functional differences⁴⁴, as well as to exhibit different responses depending on their level of compaction⁴⁵. Overall, the results highlight that the hydrogel platform induced cancer cells of various types to form spheroids in less than three days within the hydrogels, maintaining robust morphology and facilitating straightforward observation. Moreover, cancer cells are prone to remodel the ECM in the organism and spread rapidly, but they typically do not express the enzymes needed to degrade pullulan and dextran, and do not adhere to the matrix. This helps create spheroids with minimal cellular spreading over the pattern and facilitates long-term culture. Thus, the polysaccharide-based platform proved effective for spheroid production.

As a proof-of-concept, we evaluated the response of A673, MCF7, and U87 to DOX and PTX, finding similar responses in 2D and 3D cultures. In particular, MCF7 spheroids were more sensitive to PTX than their 2D counterparts. The response varied for DOX, with A673 being more sensitive in 3D, whereas U87 and MCF7 showed enhanced resistance in a spheroid configuration as compared to monolayers. Most studies in the literature show that cancer cells are more resistant in 3D. It is thought to originate from the dense cells and ECM organization in spheroids that makes drug diffusion more challenging⁴⁶, combined with a possible switch in drug resistance promoted by the 3D organization^47,48. Still, other studies showed that 3D cells could be more sensitive than their 2D counterparts^49,50. Indeed, Gomes et al. cultured MCF7 and MDA-MB-231 on Matrigel to form 3D multicellular structures and found that these spheroids were more sensitive to low doses of DOX because of the inhibition of autophagy⁴⁹. A possible mechanism of drug resistance induced by protective autophagy has already been reported⁵¹, which could be regulated by cellular organization and matrix composition⁴⁹. In this respect, Chipurupalli et al. showed that autophagy was induced less efficiently in 3D conditions with their setup, which resulted in an increase of mortality in 3D as compared to 2D⁵⁰. On the other hand, a study showed that spheroids were more resistant to DOX when grown on Matrigel versus PuraMatrix, which was explained by the lack of endogenous proteins provided by PuraMatrix⁵². A similar phenomenon could be at work for PuD hydrogels as compared to the endogenous ECM in vivo. Additionally, it might be useful to investigate the relative diffusion kinetics of the drugs inside both the hydrogel and the spheroids. For example, a retention effect of DOX within the hydrogels has been evidenced (Supplementary Fig. S4A), with an associated release over several hours after the medium is renewed (Supplementary Fig. S4B). Therefore, this retention effect might increase the local concentration of drug near spheroids, potentially lowering the apparent IC50. Additionally, the plateau showing around 20% viability for high doses of DOX has also been described in the literature^47,52,53. In the spheroid, drugs diffuse both in extra and intracellular spaces^22,54. Consequently, drug diffusion in spheroids is ruled by several complex processes and might not reach all cells, resulting in an apparent higher IC50. It can be the case for DOX, that seems not to penetrate within the A673 spheroids after 2 days by looking at the confocal images (Fig. 3), although DOX is rather small (M = 0.5 kDa). Transferring A673 spheroids to a culture-treated substrate after 48h of DOX treatment confirmed the presence of viable cells (Supplementary Fig. S3C). These spheroids remained cohesive and could grow and fuse except at high doses (Supplementary Fig. S3A). Notably, the outcomes at intermediate and high doses varied significantly, and could not be predicted by their appearance in microwells (Supplementary Fig. S3B). Another mechanism of drug retention could be the rebinding that takes place when drug site occupancy is quite low, which is the case with taxanes like PTX^39,55. It is also important to specify that our protocol for measuring IC50 in 2D requires achieving the same cell density per unit area as in 3D cultures. This inevitably results in high confluence in 2D and therefore low proliferation at the time of drug administration, which might hamper the action of antimitotic drugs and partly explain the reduced drug sensitivity in 2D. Moreover, cell confluence has already been shown to affect drug response⁵⁶. Thus, comparisons between 2D and 3D cultures are not easy and emphasize the need for 3D-tailored protocols that take into consideration interactions between the drug, the assays substrates and the cells in their specific 3D organization⁵³. To this extent, the use of classical metabolic activity assays, that were designed for 2D cell culture, can be discussed⁵⁷. Resazurin as well as its reduced form, resorufin, have intrinsic diffusion properties that are affected by the presence of the hydrogel⁵³. Yet, resazurin and MTT are among the gold-standards of cell viability assays since they are easy to use and present little toxicity for cells, enabling follow-up studies on the same cells. Besides, spheroids over 500 µm diameter often present a necrotic core^58,59, which results in the secretion of cell death markers that could interfere with supernatant assays like LDH. These large spheroids also present a subpopulation of quiescent cells as well as a very dense outer layer⁴⁵, characterized by enhanced resistance to antimitotic drugs^60,61, which makes the results of viability assays difficult to interpret.

Apart from anticancer drug testing, reproducible spheroids are produced in 3 to 5 days with good throughput and are proven viable for further quantitative analyses and confocal microscopy, including bright field size monitoring which gives an easy output to monitor their growth²⁰. More work is needed on the automatization of these analyses to make them viable at a larger scale. For this, it is worth highlighting that our design has the advantage of being adaptable to a 48 well plate and having an integrated medium chamber to ease culture and staining, while being versatile and permitting to be scaled to 6-well plates, for example, with either more or bigger microwells patterning. Finally, this platform of production can be refined to integrate more TME features. As a perspective, freeze-casting could be achieved as previously reported to obtain porous hydrogels^32,62,63, to offer an innate network for cells to spread, and for integration of a supportive cell type like CAFs, MSCs, or endothelial cells in the pores.

Methods

Hydrogel solution preparation

The hydrogel solution was made by mixing the polysaccharides pullulan (34.5% w/w, 200 kDa, Hayashibara, Japan) and dextran (11.5% w/w, 500 kDa, Pharmacosmos, Denmark) with sodium chloride (54% w/w, 58.44 Da, Fisher Chemical, USA). Powders were weighted and mixed in ultrapure water (40 mL) under magnetic stirring for 2h at room temperature (RT). The solution was then kept at RT overnight to remove air bubbles before use or stored at 4°C for up to 3 months.

3D printing of the mold

The mold used for hydrogel production was printed using a commercial 3D stereolithography-printer (028J Plus HR, DWS, Italy). The design (Fig. 1B) was made on the software Inventor (Autodesk, UK). A zoom is available in Supplementary Fig. S1. The mold was printed using a commercial ink (Ultracur3D RG35B, BASF, Germany) and cleaned with subsequent baths of water and isopropanol. The mold and glass slides were plasma-treated (Pico PCCE, Diener Electronic, Germany) and silanized to enhance hydrophobicity ((Tridecafluoro-1,1,2,2-tetrahydrooctyl)trichlorosilane, abcr Gute Chemie, Germany), before being washed using water and isopropanol. The puncher and rings used to unmold the hydrogels were printed similarly.

Hydrogel molding

The hydrogel solution was prepared by adding 5 mg of FITC-dextran (500 kDa, TdB Labs, Sweden) to 10 g of hydrogel solution under magnetic stirring at RT. After complete dissolution, 1 mL of sodium hydroxide (10 mol/L, Sigma-Aldrich, USA) was added to the mixture for 5 more minutes. In parallel, 3% w/v sodium trimetaphosphate (STMP, 305.9 Da, Sigma-Aldrich, USA) were dissolved in ultrapure water and 1 mL was subsequently added to the hydrogel preparation. The resulting mixture was rapidly homogenized and poured within the mold after 30 s of stirring. The mold was then covered with a silanized glass slide and put at 50°C for 20 min. After complete crosslinking, the hydrogels were brought under a laminar flow hood and unmolded using our custom-made setup (Fig. 1B). This setup allowed us to unmold them with great efficiency (7.5% loss on average) and transferred them with little handling to a Petri dish (Falcon, Corning, USA), where they were immersed 40 by 40 with 50 mL of PBS 10X (Gibco, USA) for 30 min twice to bring the pH back to 7–8. They were then carefully transferred to a 6 well plate (6–7 hydrogels per well) using a tweezer and a spatula and washed twice with 5 mL of sterile water. The conductivity of the water was monitored and should reach 10–15 mS at the end of the second bath (Conductimeter 145 A + , Thermo Scientific Orion, USA). Finally, 5 mL of PBS 1X (Gibco, USA) were added to each well, and the well plates were stored at 4°C until further use. Before use, the hydrogels were transferred to a 48 well plate with 400 µL of PBS 1X, and UV-C (254 nm) treated for 1h. Any PBS or medium change in the wells were performed by inserting the tip between the hydrogel and the well to reduce the shear stress applied to the cells at the center of the hydrogel, and to prevent its collapse.

Cell culture

Breast cancer cells (MCF7, ATCC #HTB22, USA), Ewing sarcoma cells (A673, ATCC #CRL1598, USA) and glioblastoma cells (U87, Sigma Aldrich #89,081,402, USA) were used for this study. They were grown in Dulbecco’s Modified Eagle Medium (DMEM high glucose GlutaMAX Supplement, Gibco #61,965–026, USA) supplemented with 10% v/v fetal bovine serum (FBS qualified, Dutscher #S1900-500C, France) and 1% v/v penicillin–streptomycin (Gibco #15,140–122, USA) at 37°C under 5% CO₂ in humidified atmosphere. Cells were cultured in tissue culture-coated flasks (TPP #707,503, Switerzland) and passaged every 2–3 days using trypsin (TrypLE, Express 1X, Gibco #12,605,010, USA). All cells were passage 35 at most, and regular mycoplasma PCR detection assays were carried out.

Spheroid formation

For spheroid formation, cancer cells were detached and counted manually in Malassez cell chambers, centrifugated at 270 g for 3 min and resuspended to the desired concentration in DMEM. MCF7 and U87 solutions were prepared at 600,000 cells/mL and A673 at 300,000 cells/mL. After UV-C treatment, PBS was removed from the 48 well plate containing hydrogels, and 100 µL of cell solution were loaded inside the hydrogel reservoir. The hydrogels were incubated for 30 min at 37°C, then centrifugated at 270 g for 3 min, and incubated for 30 min again. They were imaged using a plate reader (EnSight, Perkin-Elmer, USA) z-stack acquisition to select the picture with the best focus manually, and 400 µL of fresh medium were gently added before incubation at 37°C, 5% CO2.

Metabolic activity of spheroids treated with anticancer drugs.

After 3 days of maturation, spheroids were treated with anticancerous drugs doxorubicin (DOX, 10 mM in DMSO, Prestwick Chemical, France) and paclitaxel (PTX, 6 mg/mL in ethanol, Accord Healthcare, UK). A plate map containing 6 wells “control” containing cells treated with medium only, 3 wells “blank” that received neither cells nor drug, and triplicates or hexaplicates of 10 different doses of drug was generated (Supplementary Fig. S5). Solutions of 800 µM PTX or 100 µM DOX were prepared from the stock solution to constitute the highest dose, and each consecutive dose underwent a 4 × dilution in DMEM as compared to the previous one. The resulting range assayed was [100 µM: 0.4 nM] for DOX and [800 µM: 3 nM] for PTX. After three days of culture, hydrogels were photographed using the plate reader before the medium was removed, and 400 µL of the drug solution were added instead. For control and blank wells, complemented DMEM was used. After 48h of drug solution incubation, 40 µL of Alamar Blue 10X (Invitrogen #DAL1100, USA) were added to each well and gently mixed. The plates were then incubated for 2 to 4h at 37°C before 200 µL of each well were transferred to a 96 well plate to conduct fluorescence measurements, using a plate reader with an excitation wavelength of 560 nm and an emission wavelength of 590 nm.

Viability of spheroids treated with anticancer drugs

Alternatively, spheroids were treated with anticancer drugs doxorubicin (DOX, 10 mM in DMSO, Prestwick Chemical, France) and paclitaxel (PTX, 6 mg/mL in ethanol, Accord Healthcare, UK) after 3 days of maturation. Then, each 48 well plate were used for one drug and one cell line, with 10 doses in triplicates starting from 800 µM (for PTX) or 25 µM (for DOX) and undergoing a fourfold dilution between two successive doses. After three days of culture, hydrogels were imaged using the plate reader before the medium was removed, and 400 µL of the drug solution were added instead. For control and blank conditions, complemented medium was added instead. After 48h of drug solution incubation, 100 µL of supernatant were transferred to a 96 well plate, before 15 µL of lysis solution (LDH Cytotoxicity Detection Kit, Roche, Switzerland) were added to each well and gently mixed. The plates were then incubated for 1h at 37°C before 100 µL of each well were transferred to the same 96 well plate. The reagent mix was prepared using the catalyst (after reconstitution in milliQ water) and dye solutions, and 100 µL were added to each well (before and after lysis). After 30 min of incubation at RT, 50 µL of stop solution were added to each well before absorbance was measured using a plate reader at 490 and 692 nm.

Metabolic activity of monolayers treated with anticancer drugs.

For 2D drug screening experiments, cells were seeded in 96 well plates by adding 200 µL per well at respective concentrations of 100,000 cells/mL for A673 and 200,000 cells/mL for MCF7 and U87. After 2 days of maturation at 37°C, 5% CO₂, the medium was replaced with 200 µL of DOX or PTX solution. A well plate comprised 12 wells “control”, 4 wells “blank” and tetraplicates of 20 different doses (Supplementary Fig. S5). The highest dose was the same as in 3D treatments, and each dose was diluted twice as compared to the previous one. The resulting range assayed was [100 µM: 0.2 nM] for DOX and [800 µM: 1.5 nM] for PTX. After 2 days of incubation, 20 µL of Alamar Blue 10X were added to each well and gently mixed. After 2 to 4h of incubation at 37°C, the well plates were imaged using a plate reader with an excitation wavelength of 560 nm and an emission wavelength of 590 nm.

Viability of monolayers treated with anticancer drugs.

Alternatively, for 2D drug screening experiments, cells were seeded in 96 well plates by adding 200 µL per well at respective concentrations of 100,000 cells/mL for A673 and 200,000 cells/mL for MCF7 and U87. After 2 days of maturation at 37°C, 5% CO₂, the medium was replaced with 200 µL of drug solution. A well plate comprised 4 wells “control”, 2 wells “blank” and duplicates of 20 different doses. The highest doses were 800 µM and 25 µM for PTX and DOX respectively, and each successive dose was undergoing a twofold dilution. After 2 days of incubation, 100 µL of supernatant was transferred to another 96 well plate, before 5 µL of lysis solution were added to each well and gently mixed. The plates were then incubated for 30 min at 37°C and 100 µL of each well were transferred to the same 96 well plate. The reagent mix was prepared using the catalyst (after reconstitution in milliQ water) and dye solutions, and 100 µL were added to each well (before and after lysis). After 30 min of incubation at RT, 50 µL of stop solution were added to each well before absorbance was measured using a plate reader at 490 and 692 nm.

Determination of IC50 values

For Alamar Blue experiments, fluorescence intensities I were normalized as follow:

where is the average of the fluorescence intensities of all blank samples and is the value of the plateau of the 4-parameters sigmoidal curve that best fitted the data for each experiment in Prism 9.3.1 (GraphPad, USA). The curve was generated using a least square method with outlier removal (Q = 1%). Bar plots with statistical differences between 2D and 3D conditions are shown in Supplementary Fig. S6. The cleaned data (without outliers) were then pooled and used to generate a fit using a least square method with no outlier detection and a 4-parameter sigmoidal curve. The same method was applied to LDH experiments to compute IC50 based on cell viability:

where is the average absorbance of samples without cells nor drug at 480 nm before lysis. Values were then normalized using the value of the plateau for the best fit sigmoidal curve and all data were pooled together.

Doxorubicin dosing in hydrogels

Hydrogels were disposed and UV-treated in 48 well plates before use. They were then seeded with 100 µL of 300,000 cells/mL for 1h, centrifugated at 200 g for 3 min, then 400 µL of complemented DMEM were added. After 3 days of incubation, the medium was renewed with 400 µL of 100 µM DOX for all samples except control that were filled with an equal amount of medium. For the doxorubicin retention experiment, supernatant from two different hydrogels was recovered 1 h after adding the drug, and each day that for 3 days. On top of that, hydrogels treated for 2 days were also replenished with fresh medium for 24h before the supernatant was taken as well. For the sustained release experiment, four hydrogels have been treated with DOX, and the supernatant was recovered after 1h, replaced with 400 µL of fresh medium, and the supernatant was taken every day in the same wells until day 3, before being replenished again and being taken at day 7. Four controls without cells but with doxorubicin was also realized, whose supernatants were recovered after 1h of incubation. Finally, a calibration curve of DOX in completed DMEM was realized starting from 100 µM and doing twofold dilutions to cover the whole range of concentration assessed in drug screening experiments. The wells were incubated for 48h at 37°C, before the supernatant was recovered for absorbance and fluorescence evaluation. The plate was gently shaken before being read using a plate reader (VarioSkan LUX, ThermoScientific™). Absorbance was read at 480 and 560 nm, and fluorescence was evaluated with an excitation at 480 nm and an emission at 600 nm. Excitation and emission spectra were conducted in most concentrated samples to check any interferences with phenol red before acquisition. The calibration ladder was then realized by associating the fluorescence data with a 4-parameter sigmoidal fit, that best represented the data at low and at high doses, which shows the limitations of fluorescence linearity over such a wide range of concentrations. The subsequent curve follows the hereafter equation: with Bottom = 2.5, Top = 50.1, IC50 = 17.8 and Slope = 1.4.

Growth of spheroids treated with doxorubicin

Similarly to drug screening experiments, A673 cells were seeded at a concentration of 300,000 cells/mL in 100 µL per well in UV-treated hydrogels disposed in 48 well plates. After 3 days of incubation, a third of the samples were treated with 50 µM DOX, a third with a 16-fold diluted dose corresponding to 3.125 µM DOX, and a third were filled with fresh DMEM. After 2 days of incubation, spheroids were resuspended using a pipette and transferred to a 96 well plate, either on normal tissue culture-treated wells, or wells treated with a non-adhesive substrate (Anti-Adherence Rinsing Solution, Stemcell Technologies #07,010, Canada) for 30 min at 37°C and rinsed twice with PBS 1X. Spheroids evolution were imaged for 5 days and medium was changed at D2 after replating, after the image acquisition (Supplementary Fig. S3).

Sample staining

Cell samples were fixed using 4% paraformaldehyde (PFA, ThermoScientific #J61899, USA) for 1h at RT. After three baths of PBS 1X, permeabilization was realized by adding 100 µL of 1% Triton X100 (Sigma-Aldrich #93,443, USA) solution in PBS for 20 min at RT. The samples were then blocked with 100 µL of 2% bovine serum albumin (BSA, Sigma-Aldrich #A2153, USA) + 1% Triton X100 in PBS for 2h. They were rinsed thrice with 1 mL of PBS 1X, then 100 µL of a solution of 1:400 phalloidin (Alexa Fluor 555 phalloidin, ThermoFisher #A34055, USA) in blocking buffer was added to each sample overnight at RT. After three rinsing with 1 mL PBS 1X, the samples could be imaged or stored at 4°C in alum up to convenience.

Confocal microscopy

Images of polysaccharide hydrogels were acquired in a glass-bottom Petri dish (high glass bottom µ-Dish 35 mm, Ibidi #81,158, Germany) with a confocal microscope (LSM 780 AxioObserver, Zeiss, USA), by exciting the FITC-dextran contained in the formulation at 488 nm and using a dry Plan-Apochromat 10x/0.45 M27 optical. Images of spheroids were acquired by removing spheroids from hydrogels after staining to image them in a glass-bottom Petri using a dry EC Plan-Neofluar 20x/0.5 Ph2 optical. DOX was excited at 488 nm with a laser intensity of 20% and a gain of 800, and observed between 650 and 700 nm, whereas Alexa Fluor 555 was excited at 561 nm with a laser intensity of 5% (except for the 100 µM-treated spheroid, for which the laser was raised to 20%) and a gain of 800, and observed between 569 and 641 nm. Acquisitions of z-stacks were realized by choosing an interval of 5 µm. The images were analyzed with the software ImageJ (Fiji, USA). Colors were modified before merge, especially for DOX at low doses and actin at high doses that were increased in intensity to help visualization. For well dimensions, z-stacks were reconstituted to build 3D images on which microwell dimensions were manually measured on 3 hydrogels from 3 different batches (n = 12). For well radius, all images were analyzed using ImageJ as follows, followed by an in-app particle analysis (n = 162). The following treatment sequence has been applied:

• Convert to RGB.

• Enhance contrast (0% saturated pixels).

• Noise/remove Outliers (radius = 10 pixels, threshold = 1, which = dark).

• Make binary + Erode + Dilate + Erode + Fill Holes.

• Noise/remove Outliers (radius = 10 pixels, threshold = 1, which = dark).

• Noise/remove Outliers (radius = 50 pixels, threshold = 1, which = bright).

• Analyze Particles (size = 10,000–100,000 µm, circularity = 0–1, show = Outlines).

• Manual removal of wells that are not spherical.

Author contributions

ELV performed most experiments under the supervision of CW and TSY. CW and TSY conceptualized the study and conducted some experiments. ELV prepared the figures and wrote the manuscript under the supervision of CW and TSY. CW and TSY reviewed and edited the manuscript. All authors approved the submitted version.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-87896-7.