Plasticity of Expression of Stem Cell and EMT Markers in Breast Cancer Cells in 2D and 3D Culture Depend on the Spatial Parameters of Cell Growth; Mathematical Modeling of Mechanical Stress in Cell Culture in Relation to ECM Stiffness

1. Introduction

Most of the current approaches to cancer research and preclinical drug testing relies on 2D monolayer cell cultures, but there is a growing understanding that the other models are needed to more accurately replicate actual response of cells within the organism, including different markers’ expression and response to drugs [1]. 3D cell models were developed to mimic the structural and functional complexity of in vivo tissues. Since the information acquired from 2D monolayer culture is still a cornerstone for the research, it is necessary to establish if there are differences in expression of crucial markers between 2D conditions and more advanced 3D models.

There are different types of 3D cell culture, ranging from floating spheroids that rely on the self-aggregation of cells in a scaffold-free conditions to different types of scaffold-based spheroids, including cells cultured in specific hydrogels on standard plates or specialized 3D culture platforms [2,3]. Bioprinted 3D models (especially multicellular) seem to be the closest approximation of the real tumor [4]. It was reported that in 3D conditions the expression of stem cell markers is higher [5,6], including the research in MCF7 cell line [7], however that effect might be caused not by the induction of their expression, but by the selection of stem-cell phenotypes during spheroid formation. Epithelial-mesenchymal transition (EMT) markers were also reported to change between 2D and 3D conditions [5,8,9]. These differences can be crucial for cancer progression and metastasis, and need to be clarified. Results presented here have confirmed higher expression of stem-cell markers in 3D conditions, but the expression of epithelial marker E-cadherin displayed intriguing variability depending on the 3D model. Since there is an obvious difference in mechanical stress imposed on cells forming free-floating spheroids and cells in a culture on a scaffold, we addressed this difference by creating a mathematical model of the bioprinted culture. This model was used to compare stress fields generated in relation to ECM stiffness (different Young’s modulus, mimicking different 3D systems), to substantiate our interpretation that differences in mechanical stress on cell-cell junctions imposed by growth conditions translate into observed high plasticity of E-cadherin expression.

2. Materials and Methods

2.1. Cell Lines

MCF7 (ATCC), T47D (DSMZ) and MDA-MB-231 (ATCC) cell lines were cultured in DMEM supplemented with 10% fetal bovine serum (Thermo Fisher Scientific) or for 3D culture conditions in 3dGRO Basal Medium (Sigma Aldrich). All cell lines were authenticated by Eurofins Genomics (Germany).

2.2. Spheroid Culture

Cell suspensions were prepared by filtering trypsin-detached cells through 40 m Cell Strainer (Sigma Aldrich) and 8x104 cells were seeded into ultra-low attachment (ULA) X-well round-bottomed plates (Corning, USA) in 2 ml of 3dGRO Basal Medium (Sigma Aldrich) and cultured for 7 or 14 days, with medium supplementation in every 2 days.

2.3. Matrigel Culture

Liquid Matrigel was defrosted on ice, pipetted into 6-well plate (616 l per well) and incubated 30 min. to solidify. Cells were filtered through Nylon, Blue Cell Strainer, pore size 40μm (Sigma Aldrich) to disperse clumps and seeded at a density of 48000 per well in 2 ml 3dGRO Basal Medium (Sigma Aldrich) containing 2% Matrigel. Cells were cultured 1 to 2 weeks, medium was replaced every 48-72h. Cells were collected by trypsinization (30 min), washed with cold PBS and analyzed.

2.4. Immunofluorescence

Cells from 3D spheroid culture were harvested in the 7th day of the culture for staining and 2D cultured cells were harvested at the same time-points at the confluency ~80%. The cells were fixed using 4% PFA for 15 minutes. The permeabilization was done with permeabilization buffer: 1xPBS, 1%BSA, 0,2% Triton-X for 20minutes. Cells were incubated with primary antibodies overnight, washed 5 times with 1xPBS and secondary antibodies were added. The staining incubation was done in the 4^oC for 2 hours. Then, cells were washed 5 times with 1xPBS, suspended in the ddH2O and smeared on the microscopic glass and mounted. Imagining was done using Zeiss Axio observer Z1 LSM 800 confocal microscope. Primary antibodies: E-cadherin,Vimentin (CellSignaling), SOX2-DyLight550, OCT-4-AlexaFluor488 (ThermoFisher), secondary antibodies: Anti-Mouse IgG-AlexaFluor488, Anti-Mouse IgG-AlexaFluor488 (ThermoFisher), Anti-Mouse IgG-AlexaFluor488, Anti-Rabbit IgG-HRP (Abcam), dyes: DAPI (4',6-Diamidino-2-Phenylindole, Dihydrochloride) (ThermoFisher), calcein, Hoechst 33342 (Sigma-Aldrich).

2.5. qPCR

Quantitative PCR was performed as described [10]. Briefly, MCF7 and T47D cells were harvested from the indicated conditions and used for total RNA preparation (PureLink RNA mini kit; ThermoFisher Scientific), followed by the treatment with recombinant DNase I (Roche). 1 g of the obtained RNA was used for cDNA synthesis using Superscript III (ThermoFisher Scientific). cDNA was quantified by quantitative PCR on an ABI Prism 7500 real-time PCR system using TaqMan Gene Expression Assays (ThermoFisher Scientific, SOX2, Hs04234836_s1, POU5F1, Hs00999632_g1, GAPDH, Hs02786624_g1). The CT method was used for calculating mRNA expression levels.

2.6. Western Blot

Cells were harvested and lysed with RIPA (150mM NaCl; 1% NP.-40; 50mM Tris pH 8,0; 0,1% SDS; 0,5% sodium deoxycholate) buffer with proteinase inhibitors. Protein concentration was measured using Bradford Assay Reagent (ThermoFisher). 20µg of protein extract was boiled with 5x loading buffer and the proteins were separated on 12% SDS-PAGE and transferred to the PVDF membrane (Merck), previously activated with methanol. The membrane was blocked with the blocking buffer (5%-milk in 1xPBS) for 1h. Primary antibodies were used as listed in table 1 and incubated overnight. The membrane was washed 5 times in TBS-T buffer, incubated with secondary antibodies for 2 hours and then washed again 5 times with TBS-T buffer. Final imaging was done with the use of Mini HD 4 UVITEC system.

2.7. Biofabrication of 3D Breast Cancer Models

The architecture of the bioprinted 3D breast cancer models was designed using SLICER 4.0 3D software. Each 3D model was created with dimensions of 5x5x1 mm. The models were printed using a BIOX bioprinter (Cellink, Sweden) and cultured in vitro. The 3D-MCF-7 model consisted of three layers of the MCF-7 cell line, while the 3D-MDA model also comprised three layers of the MDA cell line. Both cell lines were used at passage 7. Printing was conducted in 24-well plates (Thermofisher, Costar) at a temperature of 22–25°C, using a 22 G nozzle with a thickness of 0.40 mm and 60% rectilinear infill. The pressure ranged from 18-20 kPa. Cell suspensions were mixed with BIOINK hydrogel (Cellink) in a 1:1 ratio and extruded into the desired patterns. All 3D constructs were stabilized by crosslinking with a 50 mM CaCl₂ solution for 5-10 minutes. After crosslinking, the DMEM medium was added, and the constructs were cultured under standard conditions. The next day, the medium was replaced with fresh DMEM, and the constructs were incubated under standard culture conditions. Bioprinted constructs were cultured for 8 weeks. 3D models were evaluated morphologically and 5 out of 34 were selected for imaging. Live cells were stained by incubation with calcein (Thermo Fisher Scientific, working solution: 10 μM 2h) and Hoechst 33342 (Thermo Fisher Scientific, 1:1000, 10 min). The imaging was performed using Zeiss LSM800 confocal microscope. Images of 54, 199, 241, 279 and 253 focal planes of the spheroid culture were generated and evaluated for modeling.

2.8. Mathematical Model

2.8.1. Methods

The cells and the extracellular matrix (ECM) were modeled using a finite element method [11]. The modelling is done with Abaqus program https://discover.3ds.com (last access 12.12.2024). The geometrical model is prepared using GiD program https://www.gidsimulation.com (last access 12.12.2024). In the analysis, the nonlinear geometry is taken into account [12,13]. This is because of the presence of the prestressing forces in the cytoskeleton [14,15]. Therefore, the nonlinear part of the strain tensor is included. Since the incremental solution is used in the Abaqus program, the strain increment reads:

(1)

where Δe and Δη are the linear and nonlinear parts of the strain increment, respectively. They are of the form:

(2)

where Δu and ∆u’ are the displacements increment vector and vector of the increment of the displacement derivatives with respect to Cartesian coordinates, respectively. The displacement vector components are (u,v,w). The symbols A, A_L stand for the linear and nonlinear operators, respectively, as follows:

(3)

(4)

The results of the analysis are the displacement and stress fields. In the results section, the displacement fields are shown. In the case of cytoskeleton, the uniaxial stress are presented. Then, the Huber-Mises-Hencky stress is given showing an effort of the material. The HMH stress reads:

(5)

where are the elements of the stress tensor describing 3D state of stress.

The solution was performed in the Abaqus program in two steps. The first step concerned the evaluation of the stress state after introducing the prestressing forces in the cytoskeleton. The second step was applied in order to impose the external pressure load.

2.8.2. Geometry

The geometry of the systems is shown in Figure 1. The cells are embedded in ECM. A simplified model of the environment is adopted; ECM is modelled as an elastic, nearly incompressible medium. ECM for a single cell is discretized with 2,248,103 tetrahedral elements, while ECM for the group of 18 cells is discretized with 2,070,659 tetrahedral elements (Figure 1, lower panel, right) The upper surface of the ECM box is loaded with an external pressure of 6.0 Pa. The model of the cell consists of nucleus, cytoplasm and cytoskeleton (based on Prendergast, McGarry, [16]). The current model is extended by employing it into group of cells embedded in the ECM. The cytoplasm is surrounded by a membrane (cortex). The nucleus and the cytoplasm are discretized with tetrahedral finite elements (87,733 and 12,046 elements, respectively). The membranes are modelled using triangular membrane elements (cortex 3236 membranes and nucleus membrane 1792 membranes, Figure 1, upper panel). The cytoskeleton is considered as a tensegrity structure based on icosahedron. The deformation of icosahedral shape is modelled on the observed actual shape of the cell. Prestressed tendons model the actin, while the bars model the microtubules. The model consists of 6 bars and 24 tendons (Figure 1, upper panel).

The model of a group of cells embedded in ECM consists of 18 cells (Figure 1, lower panel). The number of tetrahedra and triangles in each cell is similar to that of one cell. It slightly varies in each cell due to the meshing algorithm, which is sensitive to the shape of the entire structure.

3. Results

3.1. The Expression of Stem Cell Markers in Breast Cancer Cell Lines Increases in 3D Culture Conditions, Regardless of the 3D Model

The experiments were performed on the two breast cancer cell lines corresponding to luminal cancer (MCF7 and T47D). Standard 2D cultures, 3D floating spheroids on a non-adherent plate and 3D Matrigel cultures were compared for the expression of selected stem cell markers (SOX2/SOX2 and POU5F1/OCT4) on mRNA and protein levels. 3D cultures (Figure 2A) were collected and processed after 1 and 2 weeks of culture. qPCR analysis (Figure 2B) demonstrates an increase of both stem cell markers in free floating spheroids vs. 2D culture. Representative immunofluorescence images of both markers in free floating spheroids are presented on Figure 2C, with the quantification of the fluorescent signal in Figure 2D. Western blot analysis of these cultures is shown in Figure 1E. A similar analysis was performed for 3D matrigel culture of MCF7 cells (Figure 2F,G). The results indicate the increase in both mRNA and protein levels of both stem cells markers in 3D culture versus 2D culture, regardless of the type of 3D culture, although the increase of the expression is more evident for SOX2. Interestingly, after 2 weeks of culture these increased levels of stem cell markers have a tendency to normalize, especially in case of T47D. The increase in MCF7 is more pronounced and maintains longer than in T47D, thus, this cell line was selected for bioprinting experiments.

3.2. The Expression of the Epithelial Marker E-Cadherin in 3D Versus 2D Conditions Varies Depending on a 3D Model

The expression of the most basic EMT markers: CDH1/E-cadherin (epithelial) and VIM/vimentin (mesenchymal) was analyzed comparatively in the standard 2D cultures, 3D culture on non-adherent plate and 3D Matrigel cultures. Additionally, E-cadherin expression was assessed in a 3D bioprinted culture, which technically corresponds to the Matrigel type of scaffolded culture. The results indicate that E-cadherin expression increases in the culture on non-adherent plate (Figure 3A,B,C,D,E), but decreases in scaffold-based culture (both, standard Matrigel culture or long-cultured 3D bioprint sample, Figure 3D,E), compared to the standard 2D culture. Vimentin expression in the analyzed epithelial breast cancer cell lines it is not present and the both analyzed types of 3D culture do not induce it (for the reference the expression of vimentin is shown in MDA-MB-231 cell line). The second tested mesenchymal marker was N-cadherin, also not present and not induced in all tested cell lines (Figure 3D).

3.3. Bioprinting and Imagining of a 3D Culture for the Generation of the Model

Observed differences in the expression of E-cadherin, the main epithelial marker and adherent junctions protein, suggest that the difference might be caused by different tensions between cells in free-floating versus scaffold-based culture. The assumption is that free-floating spheroids should require stronger cell-cell contacts to remain intact, since they are subjected to hydrodynamic forces of the moving fluid and do not have the support of the scaffold. Thus, to evaluate tensions between cells we created a mathematical model of cellular interactions based on the imaging of the 3D bioprinted culture.

MCF7 cell line was used to generate a mature (8 weeks), self-organized 3D bioprint culture in alginate hydrogel. Out of 34 cultures, 5 were selected for imaging (Figure S1). Live cells were stained with calcein and Hoechst 33342 to stain live cells cytoplasm and cell nuclei, respectively. Selected image (Culture 1, Figure 4) was used for modeling.

3.4. Mathematical Model of Stress Analysis

The numerical model has been prepared with the finite element method, based on the previous reports [16]. The parameters assigned for the specific cell’s elements are listed in Table 1. To test the influence of the stiffness of the extracellular matrix on the stress state of cells two conditions were compared: 1) the assumed Young’s modulus of the ECM set as 56,200 N/m2 (stiff gel) and 2) the assumed Young’s modulus of the ECM set as 30 N/m2 (soft gel). The Young’s modulus of the stiff gel is significantly higher than that of the cytoplasm and the nucleus, while the Young’s modulus of the soft gel is lower than for the cytoplasm and the nucleus.

3.4.1. ECM Stress

The cell in ECM is interpreted as an inclusion in the continuous medium. The maximum HMH stress is close to the top and bottom of the cell (inclusion). In case of the stiff gel, the region of high stress is concentrated around the entire cell, similar concentration is observed for soft gel, but the stress gradients are milder. To enhance the differences present in linear scale (Figure 5A,C), decimal logarithm scale is used (Figure 5B,D). Figure 5 presents HMH stress distribution in ECM around one cell calculated for the stiff (Figure 5A,B) and soft gel (Figure 5C,D). The stress is higher in the case of the stiff ECM than the in the case of the soft ECM.

HMH stress distribution in the vertical cross-section of the matrix for a group of cells is shown in Figure 6. The maximum stress is higher in the stiff matrix than in the soft one. When comparing the single-cell and group-cell cases, the maximum stress in ECM is lower in the single-cell. This statement is valid for both kind of ECM materials, stiff and soft ECM. The maximum HMH stress is given in Table 2.

3.4.2. Cortices

HMH stress distributions in the cortex of a single cell case and the cortices of the group of cells have been calculated. The HMH stress field in the single cell in the case of stiff ECM is qualitatively different from the field in soft ECM (Figure 7). HMH stress is distributed relatively smoothly in the case of the single cell immersed in the stiff gel. The highest stress is 2.16 times higher than the lowest, while in the case of the soft ECM, the highest stress is 8.25 times higher than the lowest. The stress in the case of the soft ECM characterizes high punctual stress concentrations. They are located around the points where the cytoskeletal nodes are attached to the membranes.

The maximum stress in the cortex is higher for the soft gel, in both the single-cell case and the group-of-cell case values (Table 3). However, the qualitative picture of the stress distribution is similar. The stress concentrations are around the points of attachment to the cytoskeleton. The stress distribution is relatively smooth within the rest of the surface (Figure 8A,B).

Figure 8C illustrates the interaction between cells via ECM. The figure is prepared by uncovering the inner cells. For clarity, the cells are not touching each other, with a small gap between them. The decimal logarithm scale is used to enhance stress concentrations. Stress concentrations due to attachments of the cytoskeleton in the form of small dots are visible in both cases of ECM. However, in the case of stiff ECM, high-stress islands are visible close to the equator of the cells. In the case of soft ECM, stress gradients are lower, and the stress field is much smoother.

3.4.3. Cytoplasm

As in the cortices, the maximum stress in cytoplasm is higher in the case of soft EMC than in the case of stiff ECM (Table 4). Qualitative comparison of the stress fields in the single cell revealed that in case of both ECMs, the spots of stress concentration appeared in the sites of attachments of the cytoskeleton (Figure 9A). The vertical cross-sections of the cell show that in the case of stiff ECM, the stress concentration appears around the nucleus, which plays the role of inclusion into the cytoplasm. In the case of soft ECM, stress is significantly higher than in the case of stiff ECM, but the stress concentration around the nucleus is milder than in the case of stiff ECM (Figure 9B, enhanced in Figure 9C ).

Figure 10A,B show stress distribution in a cytoplasm of a group of cells. Stress fields in nuclei demonstrate stress concentrations in places of cytoskeleton attachments. This is similar to the single-cell case. However, the stress distributions are qualitatively different considering the case of stiff and soft ECM. A strong interaction via ECM can be noted in the case of stiff ECM. High stress is seen in the outer surfaces of the cytoplasm (Figure 10B) and close to the surfaces of the cytoplasm (Figure 10C). The stress distribution is different in soft and stiff ECM (Figure 10B). The horizontal cross-sections show high stress close to the centers of the cytoplasm (Figure 10C), opposite as in the stiff ECM case. The cells interact with each other since stress fields in cytoplasm are not homogenous.

3.4.4. Nuclei Membranes

HMH stress in nucleus membrane in the single cell case is depicted in Figure 11A. The maximum HMH stress is significantly higher in the case of soft ECM than in the case of stiff ECM (Table 5). The HMH stress distribution in the membranes is qualitatively different in the case of stiff ECM from the case of the soft ECM. In the case of stiff ECM, HMH stress is lower close to the equator of the cell than close to the poles. In the case of soft ECM, the stress is low close to the poles.

For the group of cells, the HMH stress in the nuclei membranes (Figure 12) is significantly lower in the case of stiff ECM (upper panel) than in the case of soft ECM (lower panel). The distribution of stress in the case of stiff ECM is similar in both layers of the cells. In contrast, in the case of soft ECM, HMH stress is higher in the lower layer of the cells than in the upper one (Figure 12).

3.3.5. Nuclei

The stress fields in the nucleus in the single-cell case are shown in Figure 11B. The maximum HMH stress is much more prominent in the case of soft ECM than in the case of stiff ECM (Table 6). The stress distribution is qualitatively different when comparing the two cases (Figure 11 upper panel vs. lower panel). In the stiff ECM, higher stress is present at the bottom of the nucleus, while in the soft ECM higher stress is at the upper part of the nucleus. The stress distribution in the vertical cross-sections confirms the latter.

HMM stress distribution in the group of nuclei in the cells is depicted in Figure 12. The maximum stress is lower in the case of stiff ECM (upper panel) than in the case of soft ECM (lower panel). The qualitative evaluation of stress distribution leads to the same conclusion as the estimation in the nuclei membranes.

3.3.6. Cytoskeletons

The displacements and stress distribution on the cytoskeleton in the single cell is shown in Figure 13. The displacements are significantly higher in the case of the soft ECM than in the case of the stiff ECM (Figure 13A). This is mostly the motion against z-direction. The uniaxial stress in the cytoskeleton in the single cell is shown in Figure 13B. The absolute value of stress is higher in the case of the soft ECM than in the case of the stiff ECM (minus sign means compressive stress, plus sign means tensile stress).

Figure 14 depicts the behavior of the cytoskeletons in case of the group of cells. The displacements in the case of the stiff ECM are significantly lower than in the case of the soft ECM (Figure 14A). The displacements of the upper layer of the cells are smaller than that of the lower layer. However, in the case of the soft ECM the relative difference in the displacements is more distinct. The upper layer moves almost in homogenous manner. The movement of the lower layer is much less distinct than the upper one.

The absolute value of the tensile and the compressive stress is significantly higher in the case of the soft ECM than in the case of the stiff ECM (Figure 14B, Table 7).

4. Discussion

Different growth conditions affect cell genotypes and phenotypes. Variations in the gene and protein expression between 2D and 3D cultures have been repeatedly reported, along with differences in the response to drugs [17-20].

The increase in stem cell markers expression has been already observed in 3D conditions in MCF7 cells grown on collagen scaffolds [5,21]. The authors also reported upregulation of mesenchymal markers and downregulation of E-cadherin (CDH1) in these conditions, which was confirmed by other reports [8,22]. Interestingly, for the cells grown in a different 3D model (ultra-low attachment plate) an increase in both, SOX2 and CDH1 expression was reported [9].

Plasticity of E-cadherin expression observed in our study conforms to these previous reports, although the authors of these reports analyzed always only one 3D model and did provide a comparison of a different 3D models as we do in this report. Qi et al. [23] compared scaffold-free and scaffold-based spheroids in their report, but they analyzed NSCLC cells and compared different markers. While we do not exclude the possibility that epithelial-mesenchymal transition is engaged in a scaffold-based culture, it is obviously not a case for a scaffold-free culture on a low-adherent plate. Accordingly, since we observe dramatic differences in E-cadherin levels between the two studied 3D models (scaffold-free versus scaffold-based), 3D conditions per se cannot explain these differences, and we propose the solution based on mechanobiology.

The main difference between the analyzed 3D models consists in physical conditions of growth, mainly in the fact that on non-adherent plate cells grow in the medium, with a relative freedom of movement, while on the scaffold they rest in the stiff hydrogel and have restricted movement. We assumed that free-floating spheroids require stronger cell-cell contacts (hence: E-cadherin expression) to remain intact, since they are subjected to hydrodynamic forces of the moving fluid and do not have the support of the scaffold. To test this hypothesis, we simulated the mechanical stress within the single cell and between the group of cells, using a mathematical model, based on the confocal images of a bioprinted culture.

The model has been used to compare two different conditions: with a very stiff matrix mimicking alginate hydrogel and a very soft matrix, to approximate (although not completely) the conditions in the medium. The results for every cellular element (cortex, cytoplasm, nucleus, cytoskeleton) and for a single cell as well as for a group of cells, demonstrate that the stress, especially the maximal stress, is always higher in the conditions of the soft gel. Although there are interesting differences between stress field distribution for the stiff and the soft gel, the highest values for the soft gel, especially at the attachment sites, substantiate our hypothesis that higher stress requires stronger cell-cell junctions to preserve the integrity of the group of cells. If not for E-cadherin (or possibly other junction proteins) upregulation, these group of cells would disintegrate into single cells, which in 3D conditions would be detrimental for their survival. This may be the reason for reported upregulation of some junction proteins in circulating tumor cell clusters [24,25] and have profound significance for the metastatic process.