Brain Cancer Cell-Derived Matrices and Effects on Astrocyte Migration Free

The extracellular matrix (ECM) plays a major role in cell-cell and cell-matrix signaling during normal physiology and disease. Cancer cells secrete an ECM that is compositionally distinct from the native tissue, and this matrix directly supports cancer cell invasion and disease progression [Liu et al., 2019; Serres et al., 2014]. In glioblastoma, the most aggressive primary brain tumor, cancer-derived ECM molecules recruit and educate neural and immune cells within the adjacent tumor microenvironment to adopt immunosuppressive, pro-tumor phenotypes [Sun et al., 2019; Xia et al., 2015]. Further studying cancer ECM could therefore lead to new therapies for preventing tumor microenvironment education. It may also be possible to leverage the immunosuppressive properties for calming inflammation after injury and/or promoting tissue regeneration. Currently, cancer ECM is often studied in vitro one molecule at a time or within the complex but uncontrollable tissue environment of small animal models. An approach using cell-derived matrices may provide a better alternative for studying cell-matrix communication in vitro in a more complex, tissue-like environment. Cell-derived matrices are traditionally generated by treating confluent cultures of stromal cells, like fibroblasts, with ascorbic acid to increase inherent matrix production. Unfortunately for our purposes, ascorbic acid is toxic to cancer cells [Shenoy et al., 2018]. An alternative approach may fortunately be found in the technique of macromolecular crowding (MMC).

The inherently crowded environment within tissues is poorly reproduced by the dilute conditions of traditional 2D cell culture [Tsiapalis and Zeugolis, 2021]. Adding inert macromolecules into the cell culture medium helps recreate the crowded nature of in vivo by introducing steric hindrance, electrostatic repulsion, or both [Gaspar et al., 2019]. MMC biophysically restricts diffusion of cell-derived molecules like procollagen and proteinases for matrix assembly, thereby accelerating ECM deposition. Common MMC agents are high molecular weight polysaccharides – like Ficoll, carrageenan, and dextran sulfate – because a single molecule can exclude significant medium volume while also being relatively inert. Crowding has been used to increase production and/or deposition of ECM molecules in cultures of fibroblasts [Marinkovic et al., 2021; Shendi et al., 2019; Kaukonen et al., 2017], mesenchymal stem cells [Chiang et al., 2021; Rubí-Sans et al., 2021], and some cancer cell types [Bascetin et al., 2021; Ranamukhaarachchi et al., 2019]. This approach has been proposed for applications in screening of anti-fibrotic drugs [Cavanzo et al.,2020], modeling the fibrotic tumor microenvironment [Rubí-Sans et al., 2021], and creating engineered scaffolds for tissue regeneration [Fitzpatrick and McDevitt, 2015]. MMC protocols vary widely, and the best molecule to promote ECM deposition appears to be cell-specific. Here, we optimized an MMC approach to generate glioma-derived matrices and studied the impacts on the migration, morphology, and metabolism of astrocytes, an important cell in the brain tumor microenvironment. Understanding the effects of glioma-derived ECM (GL-ECM) on astrocytes may help identify new targets to disrupt glioma-neural cell signaling and limit tumor progression or mimic this signaling and explore the prospects of cancer ECM as an “onco-regenerative niche” [Gregory and Paterson, 2018].

GL261 is a desexualized mouse glioblastoma cell line commonly used in experimental research to model glioblastoma in vitro and in vivo [Bayik et al., 2020; Cornelison et al. 2018]. These cells were a generous gift from the lab of Dr. Jennifer Munson at Virginia Tech. The cells were subcultured and maintained using Dulbecco’s modified eagle medium with 10% fetal bovine serum. All experiments were conducted with GL261 of passage number 9–12 with passaging at 80–90% confluence. To obtain a confluent monolayer, GL261 cells were plated at a density of approximately 17,700 cells/cm² or 62,000 cells per well in a 12-well plate. MMC agents were then added the day after plating (continued below).

Human cortical astrocytes (Sciencell #1800) were used to assess the effects of GL261-derived matrix on neural cell behavior, including migration. Astrocytes are subcultured on collagen-coated plates with Astrocyte Medium (Sciencell #1801) containing astrocyte growth supplement and 2% fetal bovine serum. The flasks or well-plates are coated with collagen by adding a small volume of ∼3 mg/mL rat tail collagen (Fisher 354236), spreading the liquid to cover the culture surface, and incubating at 37°C for 30 min. The flasks/well-plates are then washed with phosphate buffer saline (PBS) and filled with culture medium prior to plating the cells. All cells used for these experiments were between passage number 4 and 8. The cells were passaged at 80–90% confluence.

For spheroid generation, the cell solution was adjusted to a concentration of 2.5 × 10⁶ cells/mL, and well-spaced 10-μL drops were pipetted onto the inverted lid of a 10 cm cell culture dish. The bottom of the dish contained 10 mL of sterile PBS to maintain hydration. The lid was then inverted onto the dish, and the cells were cultured for approximately 3–4 days until stable aggregates formed. The 3D spheroids were then transferred into individual wells of a non-treated round bottom 96-well plate for further culture.

For 2D cultures, MMC agents were added to each well the day after plating. Media contained either high-molecular-weight hyaluronan (HA; Sigma 53,747) at 0.5 mg/mL; Ficoll 70 (Sigma F2878) and Ficoll 400 (Sigma F4375) at 37.5 and 25 mg/mL, respectively; or carrageenan (Sigma C1013) at 0.075 mg/mL. These concentrations were chosen based on previous reports [Marinkovic et al., 2021; Gaspar et al., 2019; Shendi et al., 2019]. The media were replaced every other day until the end of the experiment. A similar process was used for spheroids, except the MMC media were added after the spheroids were formed (∼2–3 days) and transferred into the non-treated round bottom plate.

After 1 week of 2D culture with MMC agents, the media were removed, each well was briefly washed with water to remove salts, and the plates were placed in the freezer until ready for analysis. The composition of the ECM was assessed using quantitative matrix assays, conducted according to the manufacturer’s instructions. Collagen was quantified using the Sircol^TM Soluble Collagen Assay (Biocolor) and a pepsin solubilization step; sulfated glycosaminoglycans were quantified using the Blyscan^TM Glycosaminoglycan Assay (Biocolor), and total protein content was measured using the Pierce^TM Rapid Gold BCA Protein Assay Kit (Thermo Fisher). All samples were run in triplicate and analyzed on a BioTek Cytation 3 imaging plate reader. Data for collagen and glycosaminoglycans are presented normalized to the respective total protein content for each well to control for potential variability in cell numbers.

GL261 samples for immunostaining were fixed in 4% paraformaldehyde for 30 min at room temperature, washed with PBS, and stored wrapped in parafilm in the fridge until staining. All samples were treated with blocking buffer containing 3% donkey serum (Sigma D9663) and 0.3% Triton X-100 (Sigma X100) to permeabilize the cells and block non-specific protein binding. The following antibodies were then added in blocking buffer and incubated overnight at 4°C: goat anti-collagen I (Southern Biotech 1310-01), rabbit anti-laminin (Sigma L9393), rabbit anti-fibronectin (Cell Signaling E5H6X), mouse anti-tenascin-c (Novus Biologicals NB110-68136S), and mouse anti-CS56 (Sigma C8035). The samples were then washed three times with PBS for 10 min each, and secondary antibodies in blocking buffer were added for a 1-h incubation at room temperature in the dark. Secondary antibodies used include the following: donkey anti-goat (ThermoFisher AF488), donkey anti-rabbit (ThermoFisher AF594), and donkey anti-mouse (ThermoFisher AF647). For staining of glial cell focal adhesions, the samples were blocked as above and treated with mouse anti-Vinculin clone hVIN-1 (Sigma V9264) followed by secondary antibody donkey anti-mouse AF647 (ThermoFisher). Alternatively, samples were stained with rabbit anti-GFAP (Abcam ab7260) followed by donkey anti-rabbit AF594. The samples were counterstained using DAPI (ThermoFisher) and imaged on an EVOS^TM M5000 Imaging System (Invitrogen). The image settings were kept constant for all trials to enable fluorescence quantification. Image analysis was performed using ImageJ (National Institutes of Health) to measure mean gray value (or integrated density for GFAP) of the entire field of view. Matrix samples were imaged at 4×; samples for nuclei counting by DAPI were imaged at 10×; and samples of GFAP and vimentin staining were imaged at 20×.

GL261 cells were plated and cultured for 1 week under crowded conditions with Ficoll 70/400, as described above. To isolate deposited matrix molecules for further experiments, we first decellularized the wells by treating with the pro-apoptotic agent raptinal [Palchaudhuri et al., 2015]. Wells for apoptosis were treated with 10 μM raptinal (Sigma SML1745) and incubated for 1.5–2 h at 37°C until all the cells detach from the well-plate. The well-plates were then washed three times with PBS for 10 min each followed by a 2-day wash with PBS to ensure complete removal of the raptinal.

Confluent monolayers of human astrocytes were subjected to a scratch wound healing assay to assess the effects of GL261-derived matrix on astrocyte morphology and migration. Matrix samples were isolated from GL261 cells cultured under Ficoll 70/400 for 1 week and decellularized by apoptosis, as described above. 25,000 astrocytes were then plated into each well and allowed to grow until confluence in Astrocyte complete medium. Collagen-coated wells (prepared as above for cell culture) were used as controls. A 10-μL plastic pipette tip was used to create two perpendicular scratches in the astrocyte monolayer, and images of the crossing point were taken every 4 h, starting immediately after the scratch. The plates were returned to the incubator between imaging. Transmitted light images were taken on an EVOS^TM M5000 Imaging System (Invitrogen). The wound gap was manually outlined and measured in ImageJ. Cells needed to be in contact with the collective cell mass to be considered part of the wound gap front.

At the end of the astrocyte scratch assay, the culture medium was replaced with media containing 1× alamarBlue HS (Thermo Fisher A50100). The samples were incubated for 4 h at 37°C according to the manufacturer’s protocol, and the plate was analyzed on a BioTek Cytation 3 imaging plate reader. Fluorescence was measured with an excitation wavelength of 560 nm and emission of 590 nm. Absorbance was measured at 570 nm and a reference wavelength of 600 nm (for normalization). A higher reduction of the resazurin in alamarBlue to fluorescent resorufin indicates higher metabolic activity.

All statistical tests were performed using Graphpad Prism software version 9.2. Experiments with more than two groups were first compared by analysis of variance (ANOVA) followed by appropriate t tests to compare individual groups. Cell culture samples quantified by Sircol and Blyscan assay kits; immunostaining and astrocyte metabolism were compared by paired t tests. Wound areas in the scratch assays were compared using repeated measures two-way ANOVA. Wound closure rate was obtained from a simple linear regression, and a hypothesis test was used to compare the slopes of the two lines.

A schematic representation of MMC culture is shown in Figure 1a. To identify the best MMC strategy to generate cancer-derived cell matrices, we tested three polysaccharides previously reported as effective MMC agents, namely carrageenan at 0.075 mg/mL, high-molecular-weight HA (2.5 MDa) at 0.5 mg/mL, and a mixture of Ficoll 70 and Ficoll 400 at 37.5 and 25 mg/mL, respectively. These concentrations were chosen based on prior publications [Marinkovic et al., 2021; Shendi et al., 2019; Gaspar et al., 2019]. We cultured mouse glioblastoma cell line GL261 for 1 week in the presence or absence of MMC agents, and evaluated the extent of matrix accumulation using both quantitative assays and semi-quantitative immunostaining. Collagen content, as measured by Sircol assay, significantly increased in cultures crowded with carrageenan and Ficoll up to approximately 4 mg/mL compared to less than 1 mg/mL in untreated controls (p < 0.05) (Fig. 1b). For hyaluronic acid, the collagen content was too variable to achieve statistical significance for n = 3. MMC culture also significantly increased the accumulation of sulfated glycosaminoglycans in the cell-derived matrices up to approximately 800 μg/mL compared to 50 μg/mL in controls (p < 0.05) (Fig. 1c). Carrageenan is itself a sulfated glycosaminoglycan, but Ficoll and hyaluronic acid are not sulfated and would not be detected, indicating these measurements are not simply measuring residual presence of the crowding agents.

We visualized the accumulation of type I collagen in 1-week cultures using fluorescence immunostaining (Fig. 1c–f). Upon quantification, Ficoll was the only MMC condition to significantly increase collagen I staining compared to untreated controls (p < 0.05) (Fig. 1g). While significant, the increase in collagen I staining showed a relatively smaller effect than in the quantification of total collagen, suggesting other collagens like collagen type III may also be accumulating [Bascetin et al., 2021]. It is important to note the cancer cell cultures, even in the presence of MMC, did not secrete enough matrix to peel it from the culture surface like fibroblast-produced matrices [Marinkovic et al., 2021]. Additionally, the matrix proteins appear to remain localized near cells without presenting any obvious fibrillar structure (Suppl. Fig. 1). We attempted to culture for longer than 1 week to see if the matrix would continue to accumulate, but extended culture of the cancer cells inevitably led to overcrowding and eventual cell detachment (not shown). Cancer cell growth is less contact-inhibited than with somatic cells, and high-density culture has been shown to predispose cells to apoptosis [Pavel et al., 2018], suggesting this approach may be limited in the ability to create mats of cancer-derived ECM.

We expanded our immunostaining analysis to examine accumulation of other matrix molecules after 1 week of MMC culture (Fig. 2a–d). We specifically examined the accumulation of laminin, fibronectin, tenascin C, and chondroitin sulfate proteoglycans, because each of these matrix components has been associated with glioma malignancy [Liu et al., 2019; Serres et al., 2014; Xia et al., 2016; Silver et al., 2013]. Of these four, fibronectin was the only matrix molecule found to significantly increase in the presence of MMC agents after 1 week. Both carrageenan (p < 0.001) and Ficoll (p < 0.01) increased fibronectin accumulation, with Ficoll promoting the highest accumulation (Fig. 2e). Laminin staining tended to increase, but the results were highly variable across the four trials. The increase in fibronectin may not be surprising considering it was recently found to play a role in nucleating the polymerization of collagen I in crowded 3D cultures [Graham et al., 2019].

It was found that crowding agents with a negative charge and high degree of polydispersity lead to the best ECM deposition by fibroblasts [Gaspar et al., 2019]. Here, neutral Ficoll promoted the best accumulation of GL261-derived matrix, while negatively charged carrageenan and HA had minimal effects. The results for HA are particularly surprising since glioma cells would naturally be crowded by HA in the brain microenvironment. In terms of polydispersity, the molecular weight of the HA ranged from ∼1.5 to 1.8 MDa; the carrageenan, ∼450–650 kDa; and the two Ficolls are 70 and 400 kDa. The Ficoll mixture proved optimal here despite being more bimodal than polydispersed, showing that the optimal crowding conditions are indeed cell-specific.

Three-dimensional (3D) culture better recreates the in vivo environment than 2D cultures, and adding crowding agents may further enhance the culture fidelity. Chiang and colleagues recently showed that mesenchymal stem cells cultured as spheroids in crowded conditions showed increased matrix production [Chiang et al., 2021]. Cancer cells are also often grown as spheroids as a model of the tumor bulk. Therefore, we tested if culturing GL261 spheroids with the MMC agents would recreate the trends from 2D culture, focusing on accumulation of collagen I, fibronectin, and tenascin C. The spheroids were first generated in the absence of crowding agents using a hanging droplet method, because prior studies show crowding can prevent aggregation into spheroids [Bascetin et al., 2021]. After 1 week of crowded culture, qualitative immunostaining suggests MMC culture has less influence on ECM accumulation in GL261 spheroids than in 2D cultures (Fig. 3). Surprisingly, Ficoll, which was best in 2D, showed the lowest intensity of staining in 3D spheroids. More work is needed to further evaluate and understand the effects of crowding on GL261 spheroid cultures.

Our ultimate goal was to use these cell-derived matrices to evaluate the effects of glioma ECM on neural cells. Neural cells in the brain tumor microenvironment are directly exposed to glioma-derived signals, including ECM, which can stimulate pro-tumor functions. Additionally, the tumor microenvironment has been proposed as a source of “pro-regenerative” – or at least anti-inflammatory – cues [Gregory and Paterson, 2018], and cancer-derived ECM may help inform development of therapeutic biomaterials. Previous studies have decellularized cell-derived matrices and replated new cells to show the effects on subsequent cell migration and metabolic function [Kaukonen et al., 2017; Rubí-Sans et al., 2021]. We chose to focus on decellularizing the cancer-derived matrix created using Ficoll MMC, since this condition promoted the most matrix accumulation. While detergents can be used to decellularize these samples [Rubí-Sans et al., 2021], detergents may ultimately damage the ECM and cause protein and GAG removal [Faulk et al., 2014]. We therefore leveraged our prior approach to decellularize using induction of apoptosis [Cornelison et al., 2018]. We induced apoptosis in the cancer cells using the small molecule Raptinal, which promoted rapid cell detachment from the matrix. Due to the density of the monolayers, this process took approximately 2 h to achieve complete detachment of all GL261 cells (data not shown). We then washed the matrix with PBS and proceeded to plate astrocytes.

We used a scratch assay to assess astrocyte migration on the GL-ECM compared to collagen controls (Fig. 4a–h). Collagen was selected as a control because the astrocytes are maintained and subcultured on collagen-coated flasks for routine cell culture, which promotes better cell adherence than on uncoated tissue culture plastic. Scratch assays are commonly used to evaluate cell migration related to wound healing, and this assay specifically with astrocytes is commonly used as a simplified model of neural injury and glial scarring [McCanney et al., 2017]. The cells were plated on either GL-ECM or collagen controls and grown to confluency, then a 10-μL pipet tip was used to create a cross-shaped “wound” in the wells. The cultures were then washed to remove any detached cells and imaged every 4 h. We found astrocytes migrated faster on GL-ECM than those on collagen, nearly closing the wound gap line by 24 h (Fig. 4h). While no individual time point achieved statistical significance with n = 3, the overall curves of wound area were significantly different by two-way ANOVA repeated measures (Fig. 4i) (F(1,4) = 12.18; p < 0.05). We also conducted a linear regression of the wound area data to obtain the rate of wound closure, e.g. slope of the regressions, which was significantly faster on GL-ECM compared to collagen controls (F(1,38) = 20.82; p < 0.001) (Fig. 4j). Interestingly, GL-ECM seems to promote more collective migration of the cells, whereas astrocytes on collagen showed more single cell migration patterns (Suppl. Fig. 2). Collective migration is typical during tissue morphogenesis, but more work is needed to understand the role and implications of collective and/or individual cell migration in the processes of tissue regeneration and tumorigenesis.

To start to understand why astrocytes migrate on GL-ECM, we first evaluated the morphology of cellular focal adhesions on each matrix sample. Focal adhesion complexes contain multiple proteins that link the intracellular actin cytoskeleton to the ECM at sites of integrin binding [De Pascalis et al., 2018]. Vinculin is one of the primary focal adhesion proteins, and its intracellular localization can indicate the strength of focal adhesion formation and therefore matrix binding. Focal adhesions are also linked to astrocyte morphology and pathological functionals [Cho et al., 2018]. We plated cultures of astrocytes on either collagen or GL-ECM created with Ficoll MMC, then stained for vinculin and the actin cytoskeleton. Cells plated on collagen show formation of fine focal adhesions at the leading edge, as indicated by thin fibers of vinculin (Fig. 5a–b). In contrast, astrocytes on GL-ECM showed thicker bundles of vinculin both at the leading edge and in the central region (Fig. 5c–d), which may indicate less mature focal adhesions and weaker binding to the ECM [De Pascalis et al., 2018]. A similar change in focal adhesion morphology was recently found for ovarian cancer cells cultured under MMC [Bascetin et al., 2021], and these cells also migrated faster on their own matrices produced under Ficoll crowding. These results may be due to a change in matrix architecture, since prior studies have shown collagen produces tighter fibril networks in crowded 2D and 3D cultures [Ranamukhaarachchi et al., 2019; Saeidi et al., 2012]. It is also possible for the substrate stiffness to play a role in focal adhesion formation [Zhou et al., 2017]. The matrix samples were too thin to isolate and subject to testing, but we estimate an approximately similar collagen content for GL-ECM (∼0.2 mg of total collagen, including non-fibrillar collagens) and collagen controls (0.1 mg of collagen I, assuming total adsorption).

Finally, we measured the metabolic activity of astrocytes on GL-ECM. Glial cell metabolism is linked to cellular function and reactivity [Afridi et al., 2020], and the composition of the ECM influences astrocyte inflammatory responses [Johnson et al., 2015]. We found astrocytes plated on GL-ECM were significantly more metabolically active than those plated on collagen alone (p < 0.01) (Fig. 5e). This effect did not appear to be related to general astrocyte reactivity (Suppl. Fig. 3). We did observe GL-ECM samples to reach confluency faster than collagen samples. By counting the number of cells in the confluent monolayer at the start of the scratch experiment, we find significantly more cells present within the monolayer per field of view for GL-ECM samples (*p < 0.05, Suppl. Fig. 3). These data suggest the cancer-derived matrices may influence astrocyte proliferation, but more work is needed to confirm these effects. While the functional outcomes of increased astrocyte migration and metabolism are currently unknown, it will be interesting to further evaluate astrocyte phenotype on these glioma-derived matrices, especially considering the ability of tumor-associated astrocytes to promote an immunosuppressive environment [Heiland et al., 2019].

We have shown MMC is an effective technique to increase the accumulation of ECM molecules in cultures of the glioblastoma cell line GL261. The addition of either carrageenan or Ficoll 70/400 to the cell culture medium for 1 week significantly increased total collagen and glycosaminoglycan content as well as the immunoreactivity for fibronectin. Only Ficoll showed a significant effect on the immunoreactivity of collagen type I. These effects mysteriously did not translate to 3D spheroid cultures, requiring additional studies to further evaluate crowding and ECM accumulation in GL261 spheroid cultures. We selected Ficoll as the crowding agent of choice and decellularized the culture samples by inducing cancer cell apoptosis. The isolated matrix increased the migratory ability of human astrocytes in a wound healing assay compared to those plated on collagen controls. These effects may be related to alterations in focal adhesions and increased metabolic activity. Collectively, these data show cancer cell-derived matrices enable studying the composition of glioma ECM and also the effects on neural cells in the brain tumor microenvironment, which may ultimately reveal new implications for anti-cancer therapy or neural wound repair.

We would like to thank Dr. Yubing Sun and Jamar Hawkins for access to, and assistance with, the plate reader, and also Dorcas Matuwana for training with the scratch assay technique.

The work in this article did not involve animal or human subject and therefore did not require ethics approval.

The authors declare no conflicts of interest related to the content of this article.

We would like to thank the National Institute of Biomedical Imaging and Bioengineering (NIH R21 RB031435-01) awarded to R.C.C. (PI) for partially funding this work.

R.C.C. conceived the study and coordinated with R.L. to plan the experiments. R.L. and B.M.B. conducted the experiments and analyzed the data under guidance by R.C.C. R.C.C. wrote the manuscript, and all authors contributed to editing of the final version.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.