Hydrogel-Based Spheroid Models of Glioblastoma for Drug Screening Applications

Abstract

Glioblastoma multiforme (GBM) is the most aggressive brain tumor, with median patient survival of 12–15 months even after treatment. To facilitate basic research as well as treatment development, bioengineered GBM models that adequately recapitulate aspects of the in vivo tumor microenvironment are greatly needed. Multicellular spheroids are a well-accepted model in tumor biology as well as drug screening because they recapitulate many of the solid tumor characteristics, such as hypoxic core and cell-cell communication. There are multiple approaches for growing GBM cells into tumor spheroids – non-adherent plastic dishes, hanging drop, bioreactors, and hydrogels, amongst others. Suspension spheroid models offer ease of growth, uniformity, and overall lower cost, but neglect the cell-matrix interactions, while hydrogel-based spheroids capture cell-matrix interactions and allow co-cultures with stromal cells. In this review, we summarize various approaches to fabricate GBM spheroid models as well as GBM spheroid characteristics and chemotherapeutic responsiveness as a function of hydrogel matrix encapsulation and properties, in order to advance therapies.

Glioblastoma, the Tumor Microenvironment, and Current Treatments

Glioblastoma Multiforme

Glioblastoma multiforme (GBM) is a lethal disease with poor patient prognosis that represents 16% of all brain tumors.¹ GBM is the most aggressive type of brain tumor, with median patient survival of only 12–15 months. Over 23,000 new cases of central nervous system tumors are reported, and over 14,000 deaths occur annually due to GBM in the United States.² The initial GBM symptoms are non-specific, such as headaches or nausea and worsen rapidly.³ The cause for most GBM cases is not known and there is no known method to prevent the disease. While the disease can happen at any age, it occurs most often in older adults and it is more common in males than females.⁴

GBM Microenvironment

The GBM microenvironment is extremely complex, consisting of various extracellular matrix (ECM) proteins, glucosaminoglycans (GAGs) and cell types. Healthy brain tissue contains low levels of fibrous proteins such as fibronectin, laminin and collagen and high levels of GAGs, both unbound and bound to fibrous proteins and proteoglycans.⁵ The presence of GBM tumors greatly alters the ECM both in concentration and composition, with the relative concentration of ECM compared to cellular components increasing from 20% to 48% in GBM.⁵ In GBM, there is upregulation of primarily hyaluronic acid (HA), laminin, and collagen, along with other ECM components.^6,7 This increase in ECM components is due to upregulated secretion of fibrous proteins from healthy brain tissue and HA from glioma cells.^5,6 Other crucial non-cellular components of the tumor microenvironment are extracellular vesicles and soluble signals secreted from glioma cells, which trigger cell migration and ECM secretion from surrounding cells.⁸ The GBM microenvironment also includes various cell types such as glioma cells, glial support cells (astrocytes and oligodendrocytes), and immune cells.^9,10 A simplified schematic of the GBM microenvironment is represented in Figure 1.

Figure 1.

A simplified schematic of the GBM microenvironment.

The GBM microenvironment provides structural support and directs GBM cell behavior, disease progression and therapy responsiveness.¹¹ For example, presence of the proteoglycan tenascin-C correlates with angiogenesis and the progression of glioma.¹² Collagen IV, V, fibronectin, and laminin, which abound in the basolateral membrane of GBM-associated vessels, enhance cell survival, proliferation and migration.¹³ Adhesion of integrins to their ECM ligands lessens GBM response to therapy.¹⁴ In terms of mechanotransduction, increased tumor stiffness contributes to malignancy.¹⁵ and alters integrin expression, which in turn regulates GBM cell survival, proliferation, migration and invasion.¹⁶ In terms of support cells, endothelial cells in the perivascular niche secrete interleukin-8, which enhances the proliferation and migration of glioblastoma stem cells (GSCs).¹⁷ Microglia, another support cell type, produce matrix metalloproteinases (MMPs) and modulate growth and migration of glioma cells.¹⁸ Hence, mimicking key aspects of the glioblastoma microenvironment, such as biochemical and mechanical input as well as presence of support cells, is critical for the development of faithful in vitro GBM models that could lead to the discovery and development of successful therapies.

GBM Treatments

Typical GBM treatments include maximum safe surgical resection of the bulk tumor followed by chemotherapy with temozolomide (TMZ), coupled with radiation therapy.¹⁹ TMZ is currently the only approved GBM chemotherapy, and provides only a modest two-month increase in median survival.²⁰ Overall, with current treatments, median survival rate is only 12–15 months with a two-year survival rate of only 25%.²¹ Anti-angiogenesis therapy via bevacizumab (BEV) showed promising results at early clinical trials,²² but phase III trials revealed increased toxic side effects without additional overall survival.²³ The challenge with successful chemotherapy is typically the complex tumor microenvironment and difficulty passing the blood brain barrier, which can lead to chemoresistance.²⁴ To address chemoresistance of individual chemotherapeutics, combination therapies are commonly used, but GBM recurrence is still a serious issue with current treatments.²⁵ Targeted therapies are currently being researched to help improve GBM treatment efficacy, including targeting epidermal growth factor receptor (EGFR), vascular endothelial growth factor (VEGF), and other receptor tyrosine kinase (RTK) pathways.^26,27

GBM Subtypes

Perhaps one of the biggest contributors to GBM resistance to therapies is its heterogeneity. Recent large-scale genomic analyses of GBM by the Cancer Genome Atlas Research Network identified four distinct molecular subtypes of GBM: neural, proneural, classical and mesenchymal.^28,29 Proneural GBM patients had the poorest response even to the most rigorous treatment, while patients with the other three subtypes had modest survival gains. Furthermore, each subtype had different genetic mutations and expression, suggesting the need for subtype-specific and individualized treatment. This and many follow-up studies have demonstrated GBM heterogeneity³⁰ and many others have addressed the critical importance of the microenvironment on GBM response to therapy.³¹ However, it is yet to be interrogated how individual GBM subtypes respond to microenvironmental cues and how this interplay translates to therapeutic responsiveness. This is a potential new frontier in the search for GBM treatments.

Hydrogel-Based in vitro Spheroid Models for GBM

The Need for Hydrogel-Based Multicellular GBM Spheroid Models

Multicellular tumor spheroids could lead to a cancer therapy breakthrough by modeling avascular tumor nodules in the laboratory^32,33 providing a substrate for predictive high-throughput drug screening. The need for better screening has intensified as advances in genomics and organic chemistry created a backlog of candidate drugs.³⁴ Even with high-content screening, current methods frequently fail to eliminate poor candidates: roughly 90% of late-stage clinical trials fail, often due to unforeseen ineffectiveness or side effects.³⁵ Multicellular tumor spheroids could improve the predictive capacity of drug screening platforms, relieving cost and ethics concerns of human and animal trials.

Most commonly, spheroids are grown in a liquid - on non-adhesive substrates, hanging drops or in bioreactors.³⁶ While such suspension spheroids mimic many morphological, functional and mass transport features of naturally occurring tumors,³⁷ unlike natural tumors, these mimics exist in an attachment-free microenvironment which does not resemble the mechanical, physical, and biochemical properties of the native tumor ECM. To more faithfully mimic the native microenvironment, formation of spheroids in an external matrix has emerged as a promising alternative to suspension spheroids. Unlike suspension spheroids, matrix-grown ones can mimic tumor stroma, allowing tests of therapies that target the stroma, the immune system, or stromal cells.³⁸ Recent studies also demonstrate that growing spheroids in a matrix can produce more cancer stem cells,^39,40 which have been implicated as major contributors to therapy resistance and tumor recurrence.⁴¹

Tumor Spheroid Fabrication Methods

Tumor spheroids for drug screening can be formed via clonal expansion or through cell aggregation. For clonal expansion, single cells are immobilized in a hydrogel matrix such as Matrigel,⁴² polyethylene glycol (PEG),⁴³ or fibrin⁴⁴ and allowed to grow into spheroids from a single cell. These spheroids require long growth times of one to two weeks, but are relatively simple to prepare, and the matrix provides critical physical, biochemical, and mechanical cues to the cells. However, they also result in spheroids of heterogenous sizes and shapes, where spheroid size is further limited by the formation of a hypoxic core. Alternatively, tumor spheroids can be formed in hydrogel matrices through aggregation either by loading cells at high density into hydrogel microspheres^45,46 or by seeding and allowing cells to aggregate in hydrogels with pre-templated spherical openings.⁴⁷ Such methods allow better control over spheroid size and heterogeneity, while still enabling cell-matrix interactions during spheroid maturation and growth.

For fast spheroid fabrication, cells could be seeded in specialty plates that prevent cell-surface interactions and forced to aggregate through forming strong cell-cell contacts. For example, the hanging drop method puts a high concentration of cells in a drop of cell culture media suspended in the air through surface tension, where cells aggregate and form spheroids.^{48, 49} However, hanging drop plates are difficult to handle, are not amenable to robotic handling and do not allow spheroid characterizations directly in the plates. Micropatterned plates with ultra-low attachment surfaces (ULAP) have been developed for easier handling and spheroid characterizations, where cells are allowed to aggregate in the microwells.^50,51 Spheroids of uniform sizes can be produced robustly, and the spheroid size can be controlled via the microwell size. Upon formation, which happens within 24 hours, spheroid aggregates can be characterized directly in the plate, harvested and used as is, or harvested and seeded in an external matrix and allowed to grow further in the presence of cell-matrix interactions.⁵²

Spheroid Models Using GBM Cell Lines

Cell lines have been extensively used in drug screening in general and for spheroid fabrication in particular. For example, the National Cancer Institute uses a panel of 60 cells lines representing various cancers, namely NCI-60, to screen all U.S. Food and Drug Administration (FDA) approved chemotherapeutics.⁵³ Cell lines are less expensive than patient-derived primary cells and allow a high level of reproducibility and direct comparison of results between different labs. However, cells lines do not recapitulate the properties and heterogeneity of the primary tumor cells and are pre-conditioned to tissue culture plastics (TCP), thus possibly leading to altered cell-matrix interactions.

While simplistic, GBM spheroid models involving cell lines have taught us a lot about the effect of the mechanical, physical, and biochemical matrix properties on GBM spheroid characteristics and drug responsiveness. For example, Wang et al. demonstrated that U87 GBM spheroids proliferate more and show more infiltration in soft gels compared to stiff gels.⁴³ The researchers used PEG/HA hydrogels with MMP-cleavable crosslinkers and grew the spheroids for 14 days via clonal expansion. Varying the hydrogel stiffness also affected the expression of MMPs with MMP-1 upregulated in stiffer gels and MMP-9 upregulated in softer gels. MMP production is indicative of tumor progression, as invasion increases with increase in MMP production. Matrix stiffness has also been shown to affect the invasive capacity of tumor spheroids.⁴⁸ Specifically, U87 and U118 spheroids were formed and allowed to infiltrate into varying stiffnesses of gelatin/HA/PEG hydrogels (2 vs. 5 kPa). Cells preferentially infiltrated into the softer gels. Further, infiltrating cells were able to form secondary spheroids, demonstrating how cancer cells could infiltrate into brain tissue and create tumors in secondary locations. In another infiltration study, Stein et al. encapsulated U87 spheroids in collagen gels, using cells with (U87WT) and without (U87ΔEGFR) epidermal growth factor receptor (EGFR), present in malignant tumors, to model the spheroid invasive potential.⁵⁴ The authors found that the infiltrating periphery of the U87WT spheroids had higher invasion rate into the collagen gel, but the core radius of both U87WT and U87ΔEGFR grew at a steady rate of 27 μm/day. Interestingly, the core radius of U87WT decreased initially due to cells infiltrating into the collagen gels faster than the core was able to grow.

Multiple spheroid models have also been developed to understand how the surrounding matrix affects GBM spheroid drug responsiveness. For example, U87 spheroids were encapsulated in fibrin gels and allowed to infiltrate into the matrix.⁵⁵ When exposed to the chemotherapeutic atorvastatin, the infiltration capacity of the spheroids greatly decreased, and the number of apoptotic cells increased compared to the no drug control. It is important to note that while different hydrogel matrices can be used for spheroid encapsulation, the type of matrix itself could influence the cells’ chemotherapeutic response. For example, U87 and U118 spheroids formed via clonal expansion within three to five days were treated with TMZ and Carmustin (BCNU).⁴² The authors specifically compared spheroids formed in chitosan/PEG hydrogels vs. ones formed in Matrigel, and noted more invasion in Matrigel but increased chemoresistance in the chitosan/PEG hydrogels. Further, the matrix presence alone, compared to suspension spheroids, can affect spheroid characteristics and responsiveness to chemotherapeutics. For example, Erkoc et al. compared suspension U373 spheroids to spheroids encapsulated in gelatin methacryloyl (GelMA) hydrogels.⁴⁹ Spheroids grown in GelMA show increased diameter and cell viability compared to suspension spheroids over the course of 14 days with protrusions from the spheroid at day seven. MMP production, which contributes to spheroid invasiveness, is also dependent on matrix presence as MMP-3 expression increases in the gel matrix and MMP-9 increases in suspension spheroids. Anti-cancer drug treatment using Digitoxigenin (Dgx) results in decreased spheroid size and increased sensitivity to the anticancer drug compared to single cells in GelMA.

Spheroid Models Using Primary Cells and Tumor Explants

Primary cells and tumor explants address some concerns associated with cell lines, primarily introducing more tumor complexity. Primary cells are not pre-conditioned to the in vitro environment prior to drug screening, and the chances of mutations are greatly decreased. Tumor explants are even better at mimicking the tumor microenvironment as the cells contain some of the native ECM proteins as well as various support cells. In addition, using patient-derived cells and explants opens the door to personalized therapy, where drug and drug combinations can be pre-screened in vitro to determine efficacy prior to patient administration. While primary cells and tumor explants provide more physiologically relevant conditions compared to cell lines, they might not always be practical as they lack the reproducibility and verification potential of using cell lines.

In one study, Akay et al. created a microfluidic chip to be used on tumor explants for drug screening.⁵⁰ Spheroids were formed in the microfluidic device on micropatterned PEGDA and treated with TMZ and BEV alone and in combination. The results showed that both drugs were effective in lowering cell viability, especially when used in combination. Of the three patient samples used, each sample had varying effects on tumor responsiveness due to the complex and heterogeneous tumor microenvironment. Another study specifically addressed the ability of tumor explants encapsulated in a physiologically relevant matrix, named VersaGel, to replicate patient drug responses.⁵¹ Retrospective drug screening was performed on five patients and the results seen for explants in the VersaGel corroborated clinical responses: three patients were responsive to TMZ, one patient was partially responsive and one patient was non-responsive. With the non-responsive patient samples, three different chemotherapeutics were investigated and gave results with two drugs giving IC₅₀’s lower than maximum blood concentration and one drug giving an IC₅₀ higher than maximum blood concentration, but patient response to the drugs was not determined.

Note that most spheroid models replicate avascular tumors and are usually cultured for up to 14 days due to diffusion and nutrient limitations. To further recapitulate the complexity of the tumor microenvironment, Ozturk et al. 3D printed a tumor organoid containing a tumor spheroid and blood vessel analogs.⁵⁶ Tumor spheroids were formed using patient-derived cells in a micropatterned plate, placed on top of a collagen hydrogel and sandwiched between two gelatin perfusion channels, with HUVEC cells perfused to mimic blood vessels, and subsequently covered with a collagen gel. The vascularization allowed for tumor spheroids to be cultured for two months, where spheroids infiltrated into the gelatin matrix. When treated with TMZ, the spheroids showed higher cell viability compared to single cell suspensions.

Conclusion and Future Perspectives

Hydrogel-grown GBM spheroids are promising tools for drug screening applications to address the growing backlog of drug candidates and the growing concerns with low drug efficacy or unexpected side effects. While there are a variety of preliminary studies describing the utility of such models and showing their ability to predict clinical outcomes, there are many venues available for future explorations. Those include studies to determine key hydrogel properties that affect spheroid responsiveness to chemotherapeutics or increasing model complexity by adding vascularization, perfusion, support cells and physiologically relevant ECM components. However, to be useful in drug screening, the complexity needs to be balanced by robustness and the amenability to high-throughput screening. In addition, while cell lines are useful for reproducibility, further research involving patient-derived cells would be useful in translational research and personalized medicine. Lastly, models need to take into account the different GBM subtypes and their specific microenvironment as they have shown differential drug responsiveness.