The goal of this protocol is to establish a 3D in vitro model to study the differentiation of cancer-associated fibroblasts (CAFs) in a tumor bulk-like environment, which can be addressed in different analysis systems, such as immunofluorescence, transcriptional analysis and life cell imaging.
Defining the ideal model for an in vitro study is essential, mainly if studying physiological processes such as differentiation of cells. In the tumor stroma, host fibroblasts are stimulated by cancer cells to differentiate. Thus, they acquire a phenotype that contributes to the tumor microenvironment and supports tumor progression. By using the spheroid model, we have set up such a 3D in vitro model system, in which we analyzed the role of laminin-332 and its receptor integrin α3β1 in this differentiation process. This spheroid model system not only reproduces the tumor microenvironment conditions in a more accurate way, but also is a very versatile model since it allows different downstream studies, such as immunofluorescent staining of both intra- and extracellular markers, as well as deposited extracellular matrix proteins. Moreover, transcriptional analyses by qPCR, flow cytometry and cellular invasion can be studied with this model. Here, we describe a protocol of a spheroid model to assess the role of CAFs' integrin α3β1 and its ectopically deposited ligand, laminin-332, in differentiation and in supporting the invasion of pancreatic cancer cells.
The tumor microenvironment is a very complex niche and extremely important for the maintenance and progression of the tumor cells1. It is formed not only by the cancer cells but also by stromal fibroblasts. The tumor cells are surrounded by a stroma that is specific and different from the stroma of normal tissues2. Laminin-332 is an extracellular matrix protein ectopically expressed in the stroma of different tumors, such as of pancreatic adenocarcinoma3. Moreover, the biochemical composition of the ECM and also its biophysical properties, such as rigidity and tension, change within the tumor bulk4. This tumor stroma, or "reactive stroma", is caused by an adaptation of fibroblasts to the neighboring cancer cells and by the recruitment of other very important players that develop a favorable and supportive environment for tumor progression. The differentiation of stromal fibroblasts results in cancer-associated fibroblasts (CAF). These cells can be identified using different markers such as α-smooth muscle actin (αSMA)5 or neural/glial antigen 2 (NG2)6.
The most suitable in vitro model to recapitulate the tumor microenvironment (TME) with CAFs is difficult to select. The method to mimic physiological parameters of the TME in a cost-efficient and reproducible way must be considered for such a model system. Within the TME, different processes, such as proliferation, differentiation, migration and invasion of the different cell types occur. These cellular processes can be performed individually with different methods. However, the experimental conditions must consider the cellular interactions with the tumor stroma ECM, since the stiffness of the substratum influences the CAF differentiation process. R.G. Wells commented on the impact of matrix stiffness on cell behavior and highlighted that cytoskeletal organization and differentiation status observed in in vitro cultured cells might be artefactual7. Different stimuli seem to be involved in CAF differentiation, including mechanical tension5,7. To avoid this, 2D soft substrates could be possible approaches for differentiation studies, as they circumvent the problem of the stiff culture dish plastic. A soft 2D surface, on which fibroblasts can be grown, can be collagen-I coated polyacrylamide gels, whereby the gel stiffness can be manipulated by the concentration of polyacrylamide and the gel cross-linker. The adhesion and formation of αSMA-rich stress fibers are enhanced in fibroblasts along with the gel stiffness8. These results stress the importance of soft substrate scaffolds for more physiological in vitro differentiation models. However, in our hands the experimental reproducibility and imaging of these gels were challenging. To overcome these shortcomings, we changed the 2D soft substrate system for a 3D spheroid model for differentiation and invasion studies. This model is more clinically relevant and, similar to an in vitro organoid, recapitulates in vivo cell-cell interactions, ECM production and deposition, as well as cell behavior9.
Spheroids are formed when cells lack a substrate to adhere to. When the cells are left without an adhesive surface, they aggregate to form a more or less spherical structure. If the spheroids are composed of one type of cell, they are called homospheroids; if composed of two or more different cell types they form heterospheroids.
Among the different methods for spheroid preparation, we perform the protocol using non-adherent round bottom 96-well plates. It is very effective with respect to the costs. Here, we produce both homospheroids of fibroblasts, CAF or CAFs lacking the integrin α3 subunit to examine the differentiation process and heterospheroids of CAFs or integrin α3 KO CAFs and pancreatic duct carcinoma cells (AsPC-I and PANC-I) to study the invasion into the surrounding matrix.
The aim for these studies was to use primary CAFs isolated from human pancreatic carcinoma biopsies. However, the biopsies to obtain the cells are scarce and for this reason, the CAFs used in these studies have been immortalized using lentivirus containing HTERT. They are called iCAFs, and their normal counterparts, primary human pancreatic fibroblasts, are termed iNFs. The human pancreatic fibroblasts and the pancreatic duct carcinoma cells, AsPC-I and PANC-I, are commercially available.
This protocol was used to study the effect of the laminin-332-integrin interaction in the CAF differentiation process. To prove specificity of this interaction and its function, inhibitor compounds were used: BM2, a monoclonal antibody that blocks the integrin binding site the laminin-332 α3 chain10, or lebein 1, a snake venom derived compound that blocks the laminin-binding integrins α3β1, α6β1 and α7β111,12.
For the invasion assay, cells had been transduced with lentivirus containing cDNA encoding either mCherry (iCAFs and integrin α3 KO iCAFs) or GFP (AsPC-I and PANC-I) to distinguish the different cell types in the heterospheroids. The transduction of the cells to immortalize them and/or to label them with fluorescent protein (mCherry and GFP) expression is described in a previous study13, that should be consulted for further information.
1. 3D spheroids as an in vitro model for fibroblast/CAF differentiation using different TGF-β1 inhibiting compounds of cell-matrix interaction
2. Immunofluorescent staining of spheroids
3. RT q-PCR of homospheroids
4. Flow cytometry analysis of integrin expression
5. Invasion assay using heterospheroids
The results of this experimental design are published in Martins Cavaco et al.13, which is recommended for further reading on the conclusions that were drawn from these experiments.
Figure 1, a representative image of an immunofluorescent spheroid, shows the immunostaining of the integrin α3 subunit of both immortalized normal fibroblasts and immortalized CAFs (Figure 1A), as well as the immunofluorescence quantification (Figure 1B) and the transcriptional levels of the integrin α3 subunit gene by qPCR (Figure 1C). This panel of results demonstrates that integrin α3 is up-regulated in iCAFs, as compared to the normal counterpart. This proved that integrin α3β1 can be considered a marker for pancreatic fibroblasts differentiation. This was also demonstrated by flow cytometry studies13.
Figure 1. Expression of integrin α3 subunit by iNFs and iCAFs. The integrin α3 subunit is upregulated by iCAFs, reflecting its potential as a differentiation marker for pancreatic CAFs. (A) Immunofluorescent staining of homospheroids of normal pancreatic fibroblasts (iNFS) compared to the homospheroids of iCAF shows an increased signal of the integrin α3 subunit in the differentiated cells. (B) The fluorescence signal of the cells in the spheroid was quantified with the Z-stack images of the 3D spheroid, using the ImageJ software. Means ± SEM of three independent experiments are shown and compared by t-test (*, p < 0.1). (C) The cells obtained from the dissociated spheroids were analyzed for their transcriptional levels of integrin α3 subunit. Means ± SEM of ΔΔCt-values as fold changes from two independent experiments are shown and compared by t-test. Although not reaching significance level of 0.1, the transcriptional levels of integrin α3-encoding mRNA are higher in iCAFs than in iNFs. Please click here to view a larger version of this figure.
Heterospheroids | ||
Cells type (passage up to 25) | Number of cells | Medium (1 part methylcellulose +…) |
mCherry-iCAFs + GFP-AsPC-I | 400 CAF + 400 pancreatic cancer cells | 1,5 parts cell MEM with 10% of FBS and 1% pen/strep + 1,5 parts of RPMI with 10% of FBS and 1% pen/strep |
mCherry-α3KO-iCAFs + GFP-AsPC-I | ||
mCherry-iCAFs + GFP-PANC-I | 1,5 parts cell MEM with 10% of FBS and 1% pen/strep + 1,5 parts of DMEM with 10% of FBS and 1% pen/strep | |
mCherry-α3KO-iCAFs + GFP-PANC-I | ||
Homospheroids | ||
Cell type (passage up to 25) | Number of cells | Medium (1 part methylcellulose +…) |
mCherry-iCAFs | 400 cells | 3 parts MEM supplemented with 10% of FBS and 1% pen/strep |
mCherry-α3KO-iCAFs | ||
GFP-AsPC-I | 3 parts RPMI with 10% of FBS and 1% pen/strep | |
GFP-PANC-I | 3 parts DMEM with 10% of FBS and 1% pen/strep |
Table 1. Cell composition of hetero- and homospheroids. The number of cells necessary to assemble the hetero- and homospheroids as well as the corresponding medium which should be used for the spheroid formation solution are summarized in the following table.
Supplemental Figure 1. Sequential steps for quantification of the immunofluorescent staining of a protein of interest in spheroids, using ImageJ. Please click here to download this file.
Supplemental Figure 2. Sequential steps for quantification of the number of invading cancer cells in the invasion assay, using ImageJ. Please click here to download this file.
To develop an appropriate in vitro model to study CAF differentiation is a challenging task. After employing different approaches, we concluded that a 3D spheroid model is the more practical, physiological and clinically relevant model, in which the interplay between pancreatic carcinoma cells with immortalized CAFs can be studied. This model prevented spontaneous differentiation of fibroblasts, due to artefactual stressors such as stiffness of the cell culture plastic, at least in short-term culture conditions (up to 48 h). Although the exact stiffness of the spheroid used in these studies was not measured, Ito et al. evaluated the stiffness of human umbilical vein endothelial cells and mesenchymal stem cell heterospheroids using a robot integrated microfluidic chip22. The average stiffness-index of the heterospheroids was measured 1 and 3 days after assembly. The size of the spheroids was different, the MSC seemed to increase their aggregation capacity, which resulted in a decrease in size over time (from 135.5 µm to 99.8 µm)21,22. Consequently, the stiffness of the spheroid also increased from 5.0 x 10-3 to 7.5 x 10-3 Pa22. However, the stiffness of the spheroids composed of a heterotypic population of CAFs and AsPC-I might possess mechanical properties and stiffness values different from the ones of homospheroids made of a single population of fibroblasts/CAFs. This is still unknown and might be different from what was described by Ito et al22. The stiffness also depends on the amount and cross-linkage of cell-produced and deposited ECM, on the different cell-cell interactions within the spheroid, on the number of cells within the spheroid or on the duration of culture.
Like any other model, the experimental design involves some variables that need optimization such as the number of cells necessary to form the spheroids, the time points of treatment and microscopic imaging. The described experimental design was optimized with the aim of having a spheroid size, which would have minimal influence in the diffusion of nutrients and oxygen, as there is a diffusion limit due to mass transport limitations9. For example, the transport of oxygen in spheroids with a size greater than 150-200 µm is significantly affected, as well as the diffusion of glucose and lactate, in aggregates bigger than 300 µm9,18. Spheroids with diameters larger than 400-500 µm usually present a concentrically layered structure consisting of a necrotic core surrounded by a viable layer of quiescent cells and an outer rim of proliferating cells9,24. To avoid the limited diffusion of compounds, and to eliminate the problem of the inaccessible spheroid core, cells were treated with TGF-β1, BM2 and lebein-1 prior to spheroid formation, when the cells are individually suspended in the spheroid formation solution. Moreover, to allow the availability of oxygen and nutrients to most of the cells in the spheroid and to prevent the formation of a necrotic core, the number of cells per spheroid was reduced to 750-1000 cells, which form spheroids with a diameter of 140-200 µm.
It is important to thoroughly mix the different cell types in the spheroid formation solution, so the spheroids are composed by approximately the same number of cells, avoiding significant size differences between the spheroids. Nevertheless, sometimes some spheroids can be slightly bigger than the other and that will result in more Z-stacks acquired in the microscope. Some spheroids might also deviate from the perfectly spherical form, but this is also taken in account when the ROI of the spheroid is delineated. These differences of size and shape are accounted in the normalized quantification values as explained in the next paragraph. Another important factor regarding the shape of the spheroid is the cell type. For example, fibroblasts and CAFs form spheroids with a nearly spherical form, while AsPC-I and PANC-I cells form a more disperse aggregate of cells.
For imaging the spheroids, cryosections of fixed spheroids can also be used, however the integrity of the spheroid may be compromised upon sectioning, depending on the cell type and the immunostaining protocol. Alternatively, the staining of the whole spheroid in its 3D structure, proved to be a simpler option. Acquiring microscopic Z-stacked pictures of the spheroid as a 3D structure takes on average 5-15 min per spheroid.
The immunofluorescent signals in images of spheroids are difficult to quantify as they represent planes of 3D structures. The method used was based on a previously described one, where the authors extrapolated and considered the spheroid to be a cell14. In this way, the fluorescence signal is corrected using Equation 1.
As background, the integrated signal density of a cell-free region of interest (ROI) was determined. The measured integrated signal density is the sum of the intensity values of the pixels within the selected ROI. Since spheroids are 3D structures, confocal images from different stacks are acquired. To process such images one can measure the fluorescent signal in each stack individually or use the sum of all stacks into a Z-projection. All the quantifications can be performed using ImageJ. The total corrected fluorescence (TCF) of the spheroids is determined as the normalized TCCF, considering the area of spheroid and the number of stacks, using Equation 2. This accounts for the varying sizes of spheroids and the discrepancy in the spherical shape. Analyzing the immunofluorescent staining of the spheroids using this method can take up to 5 min per spheroid.
Similarly, invasion of cells from spheroids into the surrounding stromal matrix can be analyzed by different algorithms. A straightforward method was previously described by Nowicki et al, in which the invasive cancer cells were counted23. In order to discriminate the invasive cells from the ones that do not invade, it is important to establish a spatial limit beyond which the cells are considered to have left the spheroid structure. Therefore, the starting point is the limit that corresponds to the perimeter of the spheroid images acquired at the initial time point (time 0), when the spheroids were embedded into the gel. The cell that invades the gel the furthest is considered the maximal distance an invasive cell can reach, and thus, it defines the outer rim of the invasion region. This method counts the invasive cancer cells that are present in the invading area, using the counting tool from ImageJ, and the ROI corresponding to the spheroid at time 0 h is added to the ROI manager and then transposed to the image of the same exact spheroid acquired 48 h or 72 h later. For this reason, it is important that the spheroid remains as close to the center of the planes as possible, during the different times of acquisition. Counting the number of cells can take up to 5 min per spheroid.
Another important consideration is to distinguish the different cell types in the heterospheroid. This can be performed using live cell fluorescent dyes, which can be problematic when the cell type has a high cell division rate or if the experiment is to be performed for long periods of time. The more robust and reliable way to distinguish the two different populations is to develop cell lines expressing fluorescent proteins such as mCherry and GFP. The delivery of the expression vectors can be performed using lentivirus, which results in stable expression of these proteins.
The authors have nothing to disclose.
We acknowledge Barbara Schedding's help in preparing the BM2 and lebein-1. We acknowledge Àgnes Noel for sharing her expertise in spheroid assays. We thank Sonja Schelhaas and Michael Schäfers for their help in handling lentiviral transfection under S2 conditions. We acknowledge Sabine von Rüden's assistence in preparing CAFs from pancreatic cancer tissue.
The research leading to these results has received funding from the People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme FP7/2007-2013/ under the REA grant agreement n◦ (316610) to J.A.E. Moreover, J.A.E. and A.C.M.C. was financially supported by the Deutsche Forschungsgemeinschaft (DFG) within the Cells-in-Motion Cluster of Excellence (EXC 1003-CiM). This project was also supported by Wilhelm Sander Stiftung (grant: 2016.113.1 to J.A.E.).
4',6-Diamidino-2-Phenylindole, Dihydrochloride (DAPI) | SIGMA-ALDRICH | D9542-10 | |
6F12 anti-Laminin β3 subunit Mouse (monoclonal) | Homemade | ||
A3IIF5 Anti-α3 integrin subunit Mouse (monoclonal) | Kindly provided by Prof. M. Hemler, Dana Faber-Cancer Institute, Boston | ||
Acetone | SIGMA-ALDRICH | 32201 | |
Albumin Fraction V – BSA | AppliChem | A1391 | |
Alexa fluor 488 Goat (polyclonal) anti-Mouse | Invitrogen | A11029 | |
Alexa fluor 488 Goat (polyclonal) Rabbit | Invitrogen | A11034 | |
Anti-laminin γ2 subunit Mouse (monoclonal) | Santa Cruz | sc-28330 | |
Anti-NG2 Rabbit (polyclonal) Millipore, AB5320 | Millipore | AB5320 | |
Anti-α-SMA-Cy3 Mouse (monoclonal) | SIGMA-ALDRICH | C6198 | |
AsPC-1 cell line | ATCC | Kindly given by prof. Jorg Haier's Lab | |
Bench centrifuge | Fisher Scientific | 50-589-620 | Sprout |
BM2 anti-laminin α3 subunit Mouse (monoclonal) | Kindly provided by Prof. Patricia Rousselle, CNRS, Lyon | ||
Calcium Chloride (CaCl2) | Fluka | 21074 | |
Centrifuge | Thermo Scientific | Multifuge 1S-R | |
Centrifuge tubes 50 mL | Corning | 430290 | |
Collagenase B | Roche | 11088831001 | |
Collagen-I, rat tail | Gibco | A10483-01 | |
Confocal microscope | Zeiss | LSM 700 and 800 | |
DMEM (High glucose 4.5 g/L) | Lonza | BE12-604F | |
Dnase I | Roche | 10104159001 | |
Flow Cytometer | BD Biosciences | FACSCaliburTM | |
Gelifying matrix | ThermoFisher Scientific | A1413202 | Matrigel, Geltrex |
Goat IgG, isotype | DAKO | X 0907 | |
Horse Serum | SIGMA-ALDRICH | 12449-C | |
Human Primary Pancreatic Fibroblasts | PELOBiotech | PB-H-6201 | |
Incubator | Heraeus | B6060 | |
Laminin-332 | Biolamina | LN332 | |
MEM | SIGMA-ALDRICH | M4655 | |
Microplate, 96 wells, U-bottom | Greiner Bio-One | 650101 | |
Microscope Slides | Thermo Scientific | J1800AMNZ | |
Mouse IgG, isotype | SIGMA-ALDRICH | I8765 | |
Multi axle rotating mixer | CAT | RM5 80V | |
PANC-I | ATCC | Kindly given by Prof. Jorg Haier's Lab | |
Paraformaldehyde | Riedel-de Haën | 16005 | |
Penicillin/streptomycin | Gibco | 15140-122 | |
QuantiTect Reverse Transcription Kit | Qiagen | 205310 | |
Rat IgG, isotype | Invitrogen | 10700 | |
Reaction tubes, 1.5 mL | Greiner Bio-One | 616201 | |
Real-time PCR cycler | Qiagen | Rotor-Gene Q | |
RNeasy Mini Kit | Qiagen | 74104 | |
Rotor Gene SYBR Green PCR Kit | Qiagen | 204074 | |
RPMI | Lonza | BE12-702F | Add glucose to 4.5 g (0.2 um filter) and 1% sodium pyruvate |
TritonX-100 | SIGMA-ALDRICH | X100RS | |
Vórtex | Scientific Industries | Vortex-Genie 2 | |
μ-Slide Angiogenesis, uncoated | Ibidi | 81501 |