Microphysiological systems to study tumor-stroma interactions in brain cancer

Abstract

Brain tumors still lack effective treatments, and the mechanisms of tumor progression and therapeutic resistance are unclear. Multiple parameters affect cancer prognosis (e.g., type and grade, age, location, size, and genetic mutations) and election of suitable treatments is based on preclinical models and clinical data. However, most candidate drugs fail in human trials due to inefficacy. Cell lines and tissue culture plates do not provide physiologically relevant environments, and animal models are not able to adequately mimic characteristics of disease in humans. Therefore, increasing technological advances are focusing on in vitro and computational modeling to increase the throughput and predicting capabilities of preclinical systems. The extensive use of these therapeutic agents requires a more profound understanding of the tumor-stroma interactions, including neural tissue, extracellular matrix, blood-brain barrier, astrocytes and microglia. Microphysiological brain tumor models offer physiologically relevant vascularized ‘minitumors’ that can help deciphering disease mechanisms, accelerating the drug discovery and predicting patient’s response to anticancer treatments. This article reviews progress in tumor-on-a-chip platforms that are designed to comprehend the particular roles of stromal cells in the brain tumor microenvironment.

1. Introduction

Brain cancers include primary brain tumors (mostly gliomas and meningiomas), brain metastases (from non-small cell lung cancer (40-50%), breast cancer (15-25%) and melanoma (5-20%) (Eichler et al., 2011)) and medulloblastoma (most common malignant brain tumor in children). They differ in the cell and tissue of origin, but they all thrive in the unique microenvironment of the brain tissue. Indeed, the brain tumor microenvironment (TME) is a vital regulator of cancer progression and therapeutic resistance (Asghar et al., 2015; Quail and Joyce, 2017). Difficult access to study human brain function has motivated transformative efforts to develop ex vivo platforms to examine the role of TME in disease progression (Figure 1). Studies have revealed tumor-stroma interactions that are unique of the brain, and the understanding of these complex associations will expand the landscape of possible therapeutic strategies. Some stromal components associated to brain tumors include the extracellular matrix (ECM), brain resident cells (astrocytes and neurons), the blood brain barrier (BBB), and resident immune cells (microglia).

Figure 1.

Graphical representation of elements that can be included within the MPS design of a brain tumor on a chip, including the ECM, the immune system (TAM and T cells), neural networks, cancer cells, astrocytes and the neurovascular unit. Tumor-associated macrophages (TAM) interact with the tumor mass, the peripheral immune system and activated astrocytes. The release of cytokines by endothelial cells and reactive astrocytes affects TAM and glioma stem cells (GSC), respectively. Created with BioRender.com.

Microphysiological systems (MPS), also referred as organ-on-a-chip platforms, are microfabricated devices bioengineered to recreate functional structures of human organs in vitro. They provide complex in vitro modeling systems that utilize multicellular structures of high physiological relevance to analyze disease pathology, progress, and response to different therapeutic agents (Osaki et al., 2018). They present key advantages such as low cost, and the incorporation of 3D multicellular cultures that recreate in vivo conditions such as flow associated shear stress and mechanical strain (Douglas et al., 2020). These devices provide the real time monitoring of key cellular parameters (Lee et al., 2020; Shin et al., 2019). MPS combine advances in microfabrication, microfluidics, biomaterials, biosensors and cell biology. A lung-on-a-chip that permitted gas exchange between endothelial and alveolar epithelial cells was one the earliest models (Huh et al., 2010). Since then, these chips have been able to mimic a multitude of organ tissues, increasing in complexity, resolution, throughput, and interactivity.

The development of in vitro cancer models must recreate the complexity of the host tissue, in order to mimic the tumor phenotype and the interaction with the surrounding healthy tissue. Tumors have high metabolic demands and surrounding vasculature is an essential element in the tumor microenvironment. A vascularized model of neuroblastoma has been described that recreates the functional vasculature and drug resistance (Villasante et al., 2017). Noncellular components of the tumor microenvironment include pH, hypoxia, and gradual diffusion of oxygen and nutrients through the tissue. Microfluidic devices have been designed to achieve spatial and temporal control over the oxygen tension, they have reported the effects of hypoxia on angiogenesis (Lam et al., 2018) and metabolic activity of tumor cells (Palacio-Castañeda et al., 2020). Ingber et al. have also described a microfluidic intestine-on-a-chip model that allows the control and monitoring of physiologically relevant oxygen gradients (Jalili-Firoozinezhad et al., 2019). MPS afford controllable microenvironmental signals to maintain physiologically relevant, tissue-specific functions.

We review here the recent advances in the design and manufacture of microphysiological systems that can recreate the particular features of some brain malignancies. We will focus on the influence of tumor stroma in regulation of cancer progression. Tumor stroma is mainly composed of the vasculature, astrocytes, neurons, ECM, and immune cells. Continuous innovation in biomaterials allows for the preparation of three-dimensional scaffolds, that combined with microfabrication techniques provide access to in vitro microtissues with highly controlled architecture, stimulating flow and continuous monitoring. These cost-effective systems can facilitate the interpretation of the roles of tumor components on brain cancer survival and invasion (Chonan et al., 2017). The applicability of these models as preclinical platforms may improve the efficacy of potential treatments and reduce toxicity.

2. Microphysiological systems in cancer research

The understanding of cancer initiation, progression and treatment resistance is essential for the design of effective therapeutic approaches (Aldape et al., 2019). The lack of significant progress in the treatment of brain diseases stimulates the design and fabrication of highly controlled preclinical platforms that investigate relevant cell interactions associated to disease pathophysiology. Due to the high heterogeneity and difficult access to human brain pathologies, research has focused on sophisticated microfluidic based microphysiological systems that recreate the dynamic native pathology of the tumor niche. In addition to the incorporation of smart biomaterial matrices into intricated microfluidic designs, research has also focused on the development of biochemical sensors for the monitoring of cell behavior and evaluation of functional structures.

2.1. Design and preparation of microdevices

Drugs targeting the central nervous system (CNS) are 45% less likely to enter Phase III trials than non-CNS drugs, according to data from 1990 to 2012 (Kesselheim et al., 2015). The current challenge in drug screening for organ specific diseases is the inaccuracy of in vitro and in vivo models. Traditional 2D and 3D in vitro models fail to properly predict native brain tissue structure, while in vivo models fail to model the involvement of the human immune system. This problem can be solved by the fabrication of organ-on-a-chip and be extended to tumor-on-a-chip systems for cancer specific treatments. Tissue chips have successfully modelled the pathophysiology of human brain tissues (Tan et al., 2021). The development of high throughput individualized brain models can be essential in the discovery of signaling mechanisms that determine the pathogenesis of brain diseases. Designing MPS of brain diseases involves the recreation of the main factors related to brain pathologies in the: device architecture, cellular and extracellular composition, tissue microstructure, flow properties, cell-cell and cell-ECM interaction and oxygen gradients, among others. These devices comprise multiple compartments and allow for the determination of biomarkers and multiple readouts.

Devices.

Multiple innovative microfabrication techniques such as micromolding, soft lithography and transfer printing have been described recently in the development of mini-tissue models (Kimura et al., 2018; Puryear Iii et al., 2020; Sontheimer-Phelps et al., 2019). Organ and tumor-on-a-chip systems have been fabricated using soft lithography (Ishahak et al., 2020), photolithography, laser printing, 3D bioprinting (Hamid et al., 2015), injection molding (Ma et al., 2020) or contact printing depending on the specific application (Wang et al., 2020). An injection-molded plastic array culture platform provided a microfluidic system for easy and versatile patterning of 3D vascular structures, integrating an established multiple-drug regime, that monitors glioblastoma (GBM) tumor response to multiple chemotherapeutic molecules (Lee, S. et al., 2019). The fabrication of the chips also encompasses a wide array of different polymeric materials including polydimethylsiloxane (PDMS), polyetherimide, polystyrene (PS) (Borysiak et al., 2013), polycarbonate (PC), polymethylmethacrylate (PMMA) (Bertana et al., 2018), silk proteins, glass, silicon, and hyaluronic acid (Whitesides, 2006). PDMS is predominant in the field due to its properties such as transparency, biocompatibility, flexibility, gas permeability and resolution. Alternative materials are chosen for their properties or applicability such as Teflon being employed for maximum chemical resistance or polycarbonate for pathogen detection (Mofazzal Jahromi et al., 2019). Alternative to PDMS, brain tumor models have been fabricated from photopolymerized polyethylene glycol (PEGDA) microfluidic systems to avoid non-specific protein adsorption, shorten production times and provide a tunable drug release capability (Fan et al., 2016).

A microchip recreating intestine–liver–glioblastoma was developed, allowing for research into how the gastrointestinal system may affect drugs that enter the body orally before they reach their targeted destination (Jie et al., 2017). The chip consisted of two PDMS layers, the bottom layer containing two olivary chambers and the top layer containing a hollow fiber. The hollow fiber was cultured with intestinal cells and the olivary chambers contained liver or glioblastoma cells. Media was passed through the hollow fiber containing intestinal cells and allowed to diffuse into the chamber containing liver cells, which was then flowed into the glioblastoma cell chambers. Overall, these configurations are important for modeling how the tumor will respond to different therapeutics while in the body and give a more accurate model for understanding the responsiveness of GBM types to specific treatments.

Biomaterials.

The recreation of the native microenvironment where brain cells reside is essential for their correct differentiation, proliferation and function. Brain tissue has a complex architecture, anisotropic and extremely soft, that makes it challenging to recapitulate (Axpe et al., 2020; Budday et al., 2017). Biomaterial scaffolds, such as hydrogels, provide versatile three-dimensional microenvironments for physiological models. Hydrogels are suitable platforms for cell culture and coatings of microfluidic channels (Annabi et al., 2013), relying on their biocompatibility and vital properties such as permeability, stiffness, and elasticity. These properties can also be adjusted during the hydrogel fabrication process to better fit the needs of the model (Ding et al., 2020). Agarose (Lin et al., 2016), collagen (Kothapalli et al., 2011), Matrigel (Kasendra et al., 2018), alginate (Zhang et al., 2016), fibrin (Jeon et al., 2015), gelatin (Paguirigan and Beebe, 2007) and cellulose (Strong et al., 2019) hydrogels have been use to reproduce the tissue characteristics of a variety of organs.

Cell sources.

An important component to any organ-on-a-chip system is the addition of cells, whose selection varies depending on the problem being investigated. Human brain cells have limited availability; therefore, brain model systems have also considered animal cell sources. Animal cells allow the collection of a high number of cells, with profiles of protein expression from the corresponding human cells. Human immortalized cells are also an alternative that is commercially available (Low et al., 2020). In the investigation of many brain diseases multiple factors can affect the molecular, chemical, and structural connections between the different types of neural cells, in addition to the difficulty of accessing diseased brain cells. In these cases, the development of human induced pluripotent stem cells (hiPSC), that can be differentiated into distinct cell types (such as astrocytes, neurons, and oligodendrocytes), has been a very valuable tool (Mofazzal Jahromi et al., 2019; Shi et al., 2017). Neural stem cell (NSCs) differentiation into both neuronal and glial cell is sensitive to physiochemical agents in the microenvironment, leading to organ-on-a-chip systems to be a prime location for culturing and evaluation of these cells. Embryonic germ cells (EGC) can also be used to further differentiate into neural cells (Mofazzal Jahromi et al., 2019; Streckfuss-Bömeke et al., 2009; Tursun et al., 2011). Brain-on-a-chip platforms have been fabricated using iPSC-derived GABAergic neurons and astrocytes with dynamic medium flow. This device provided information about neuron-astrocyte interactions that stimulates neuronal function (Liu et al., 2020).

Compartmentalization.

Microscale compartmentalization enables the co-culturing of cells in close contact and is a useful strategy to fabricate platforms that recreate the unique and essential characteristics of the tissue’s environment and the relevant structures and interactions. Compartmentalization techniques allow for the independent control of spatial, chemical, and physical culture cues. The separation of somatic cells from a glia compartment has allowed the differentiation of precursor oligodendrocytes into myelinating cells (Park et al., 2009). The organ-on-a-chip system can be extended to include multiple organs interconnected to study different complex phenomena such as drug toxicity and cancer metastasis. This is especially important in cancer research, where the evaluation of different organs, such as the gastrointestinal or immune system effect on the treatments that are targeted at the tumor site (Caballero et al., 2020). Organ-on-a-chip systems applied to the brain have more difficulties mimicking the complex and complicated organization pattern. A challenge exists in the design of brain-on-a-chip systems that lead to neurogenesis, brain growth, and defined regions such as the forebrain and hindbrain. Some advances in this area have shown that engineered neuronal-glial environments, interfaced by endothelial cells display characteristics of the blood-brain barrier, and can be applied to monitor the migration of human neural progenitors (Kilic et al., 2016). Compartmentalized microfluidic platforms have also been applied to cancer research, enabling the real-time monitoring of the formation of neuronal networks and the migration of endothelial and tumor cells under normoxic and hypoxic conditions (Gao et al., 2011). These complex microfluidic systems can also be applied to study the patient-specific responses to certain therapeutics for glioblastoma (Fan et al., 2016). The fabrication of these tumor-on-a-chip models can be improved by the use of bio-ink solutions in a concentric-ring structure that provides key tumor features such as matrix signaling and oxygen gradients (Yi et al., 2019). Tumor-on-a-chip systems have also been fabricated, containing a polyacrylamide gel, to measure both cell motility and protein expression (Lin et al., 2018).

Hypoxia.

Hypoxia, deprivation of oxygen supply, is a critical factor of the tumor microenvironment; oxygen gradients exist from the tumor core into the periphery that influence tumor growth and the effectiveness of radiation and chemotherapy (Lin et al., 2020). The development of in vitro platforms that recapitulate this environment will allow for the development of therapies that specifically target hypoxia. Recently, a tumor-on-a-chip platform, consisting of a central chamber that hosts the cancer cells, and two side media channels that block oxygen diffusion by an embedded sheet of polymethyl methacrylate, has demonstrated a metabolic switch of the tumor cells towards increased glycolysis (Palacio-Castañeda et al., 2020). Additionally, an in vitro microfluidic platform enabled series of oxygen concentrations on a single continuous microtissue (Kang et al., 2020). A polystyrene-based microdevice was fabricated by injection molding, creating a gradient of oxygen and glucose concentrations inside the device that led to the formation of a ‘necrotic core’, providing real-time data on glioblastoma cell proliferation and growth, reactive oxygen species (ROS) generation and apoptosis (Ayuso et al., 2016).

2.2. Cell tracking technologies and tumor-on-a-chip readouts

The increase in complexity of biological in vitro models has challenged the development of advanced throughput technologies. The integration of biosensors with physiological microdevices have allowed the assessment of cell activity, viability and concentration of oxygen and metabolites. The capabilities provided by monitoring multiple metrics in real time with spatial resolution are of great importance in the diagnosis of disease stages and the evolution of potential treatments. Increasing progress in electronic materials and surface modification techniques permit dynamic real-time measurements of cellular analytics and the manipulation of cellular functions. Most cell tracking techniques have focused traditionally on microscopy, not providing an integral notion of the tissue functionality. Bioelectronic platforms have been developed to provide a spatial-temporal insight into cell-cell and cell-ECM interactions (Soucy et al., 2019) and the real-time monitoring of BBB functionalities such as the trans-epithelial electrical resistance (TEER) (Brown et al., 2015). A significant challenge to record cellular activity in 3D cultures is the improvement of signal-to-noise ratio, achieved by close contact between the electrode and the cell. In this regard, micro and nano structures have been incorporated to electrodes to record neurons (Dipalo et al., 2017; Du et al., 2019). The monitoring of the activity of brain cellular networks in real time is an important requisite to understand tissue functionality.

PDMS is the most widely used material in microfluidic device fabrication, due in part to its transparency, that allows for high resolution imaging (Hachey and Hughes, 2018). This advantageous property allows for noninvasive imaging on the tumor MPS, monitoring cell fate and tissue functionality in situ. These techniques include immunofluorescence / immunohistochemistry (Lee, S. et al., 2019) and cell tagging, that can be combined with off-chip assays such as RNA-expression analysis and ELISA (enzyme-linked immunosorbent assay) (Junaid et al., 2017). MPS are appropriate platforms to investigate cancer metastasis, where multiple chambers are connected by recirculating media, and the presence of fluorescently labelled tumor cells can be detected by microscopy throughout the system (Aleman and Skardal, 2019). Tumor chips have been designed to study motility and protein expression of single cells in glioblastoma, with additional features to record programed death and viability using fluorescence (Lin et al., 2018). Moreover, a multicellular system for analyzing the interaction of intestinal and liver cells with glioblastoma cells used a removeable hollow fiber to perform scanning electron microscopy (SEM) and fluorescence imaging on the human colon cells. Additionally, fluorescence probes were used to measure the effects of the different drug combinatorial treatments on the glioblastoma cells (Jie et al., 2017).

Drug screening.

Organ-on-a-chip systems can be used to effectively develop new drugs, bringing down both the time and the costs associated with drug development and evaluating potential toxicity. In order to properly measure the interactions between organ systems when looking at drug screening, fluidic connections between these organ chambers are important parts of a multi-organ-on-a-chip. Additional variations on these multiplexed systems can be precisely tuned to measure organ-organ interaction (Sun et al., 2019). Static microscale platforms rely on direct physical contact through tissues or passive diffusions to measure the organ interactions (Diao et al., 2006). These platforms have the benefits of being easy to construct and operate, though they are subject to disturbances in gradients leading to some disparity in resulting data. Singe-pass microfluidic platforms apply an open-loop passing through the organ chambers in a unidirectional fashion, lacking some potential organ-organ interactions due to the unidirectional nature of the fluid loop (Wang et al., 2018). Neurospheroids were cultured within a chip under continuous interstitial flow and conditions to mimic either a normal brain or one affected by Alzheimer’s disease. This model was then able to model the effect of amyloid-β on the neurospheriods (Park et al., 2015).

Pump-driven recirculating multiorgan microphysical systems (MOMs) for parenchymal tissues incorporate pumps to circulate fluids between several organ chambers. Pumpless recirculating microfluidic platforms can be developed, moving away from unidirectional flow by using a combination of a rocking motion and gravity to move fluids through the chips. These MOMs can be applied for the measurements of toxicity and drug efficacy (Wang et al., 2018). Bioprinted GBM-on-a-chip models have the added benefit of using patient derived GBM cells to measure resistance to treatment of chemoradiation (CCRT) using temozolomide (TMZ). The GBM-on-a-chip model was better able to demonstrate the responsiveness for GBM cells, which was not seen in monolayer and spheroid-culture systems. There was an added benefit of being able to establish these GBM-on-a-chip systems within one to two weeks, increasing their usefulness in point-of-care testing (Yi et al., 2019). These types of systems can also model the blood-brain barrier, which normally limits the molecular exchange to the brain and access to the central nervous system. Currently, therapeutics used to target a wide array of brain diseases fail to properly penetrate the blood brain barrier, decreasing their efficiency (Wang et al., 2018). Microfluidic chips produced from soft lithography were channeled and brain and endothelial cells were incorporated into collagen gels. These models can then better model the permeability through the engineered micro vessels lined with endothelium (Herland et al., 2016). Models can combine human brain endothelial cells to better measure the efficiency of the penetration and can be paired additionally with gastrointestinal and liver cells, obtaining a more complex model of the different factors affecting drug efficiency. These models are still facing some hurdles, such as the media required for the variety of cells and phenotypes hosted within the different compartments. The materials that make up the chips are also a source of interference, skewing data by adsorbing and absorbing the added drugs (Wang et al., 2018). Additionally, the use of primary cells hampers standardization due to patient-to-patient variability and limited availability of certain cell types. Moreover, primary cells often lose their original phenotype in in vitro culture or show a limited proliferation capacity (Khalil et al., 2020).

3. Tumor – stroma interactions

Tumor progression, including brain metastasis, is profoundly influenced by elements of the tumor microenvironment such as cell populations, matrix architecture, stiffness, shear stress, chemotaxis, and hypoxia. In the context of brain tumors, we are describing here the role of extracellular matrix, astrocytes, blood-brain-barrier, neurons and microglia.

3.1. Extracellular matrix

The extracellular matrix (ECM) provides physical support to cells in tissues and organs, as well as the signaling that regulates multiple cellular functions such as growth, differentiation, and migration (Rauti et al., 2020; Theocharis et al., 2016). The ECM is also responsible for the mechanical properties of tissues such as tensile strength (Kai et al., 2019) and is made up of macromolecules such as collagen and glycoproteins, of a composition specific to the tissue and organ. Cells interact with the ECM via surface receptors that activate signaling cascades to regulate cell function. Cell – ECM interactions are highly important in carcinogenesis and tumor progression (Sood et al., 2019). Tumor cells have the ability to remodel the ECM by changing the composition, physical properties and secreting or degrading the matrix to support its survival and growth (Eble and Niland, 2019). Glioma cells alter the concentration of ECM components and the expression of the corresponding cell receptors for ECM components (Virga et al., 2019). Remodeling of the ECM can produce environments that benefit tumor cell invasions, such as the loss of basement membrane seen in some malignant transformations in breast and pancreatic cancers (Kai et al., 2019; Laklai et al., 2016). Brain ECM also regulates cell behavior and influences tumor progression, becoming an essential actor in the fabrication of physiologically relevant and instructive in vitro platforms (Wolf et al., 2019).

In healthy brain tissue, the brain ECM is responsible for cell mobility, cell communication, and regulating cell growth. This brain ECM also undergoes constant remodeling and tumor cells can manipulate those roles of the brain matrix components to accommodate its physiological needs (Bellail et al., 2004). Glioblastoma (GBM), an aggressive brain cancer, manipulates the brain ECM in order to improve its ability to survive and progress through the brain tissue. Collagen, which normally serves as a support for cell movement, is produced by GBM as a means of increasing tumor cell invasion (Chonan et al., 2017; Quesnel et al., 2020). Hyaluronic acid (HA) is the main component of the brain ECM (Figure 3), tumor cells are able to modify HA biosynthesis and is an active constituent in tumor survival and therapeutic resistance (Pedron et al., 2017; Wolf and Kumar, 2019).

Figure 3.

Brain extracellular matrix (ECM) composition varies between: (a) the basement membranes that surround the cerebral vessels and contribute to the unique structure of the blood brain barrier (together with endothelial cells, pericytes and astrocytes), (b) the hyaluronan and tenascin R based interstitial matrix between cells of the central nervous system parenchyma and (c) the tumor matrix that has increased concentration of collagen, fibronectin and laminin. Created with BioRender.com.

Multiple 3D bioengineered platform that combine microenvironmental signals of native tumor ECM have been fabricated to evaluate patient-derived brain tumor responses, such as gelatin (Pedron et al., 2019), collagen (Ozturk et al., 2020), hyaluronan (Herrera-Perez et al., 2018), silk (Sood et al., 2019), fibrin (Lee, C. et al., 2019) and decellularized tissue (Koh et al., 2018), among others. With increasing complexity, flexible extracellular matrix gels have been adapted to microfluidic devices to better mimic normal and cancerous tissue (Sontheimer-Phelps et al., 2019). Extracellular components have been introduced in microtumor models, Yi et al. showed that bioink composition affects invasion and angiogenic potential, and the sensitivity of cancer cells to treatment (Yi et al., 2019), when using a decellularized brain ECM as compared to a collagen matrix (Figure 2A). A plethora of biocompatible synthetic and natural biomaterials have been synthesized for mimicking the brain ECM, however we still lack a precise characterization of the brain tissue both in vitro and in vivo, which will have considerable implications in the development of microsystems for the analysis of brain pathologies.

Figure 2.

Tumor-stroma interactions in different microphysiological systems: (A) Illustration of a compartmentalized brioprinted GBM-on-a-chip with different bioinks. Immunostaining images show endothelial cells (CD31) form more tubule networks in the decellularized ECM (BdECM), reproduced with permission from reference (Yi et al., 2019), Springer Nature. (B) Microfluidic device (AIM Biotech) for culture of cells in a fibrin hydrogel. Cells form microvessels by day 2 (D2) and brain tumor cells migrate across the vascular network, reproduced from figure 1 a–b (Xiao et al., 2019), Wiley-VCH, published under the Creative Commons Attribution License (CC BY 4.0; https://creativecommons.org/licenses/by/4.0/). (C) Illustration of a compartmentalized microfluidic chip for studying neuron-cancer cell interactions. Immunofluorescence images show neurons and cancer cells within the device, reproduced with permission from reference (Lei et al., 2016), Royal Society of Chemistry. (D) Schematic illustration of the peripheral compartments of the MPS used to culture brain tumor cells and primary immune cells. Immunofluorescence image shows the microvessel lumen (yellow) in contact with CD8+ T-cells (green) and GBM tumor cells (red). Scale bar is 50 μm, reproduced from figure 2c and Figure 2 - figure supplement 1 A-B (Cui et al., 2020), eLife Sciences Publications, published under the Creative Commons Attribution 4.0 International Public License (CC BY 4.0; https://creativecommons.org/licenses/by/4.0/).

3.2. Neurovascular unit

The blood-brain barrier (BBB) is a crucial structure that acts as an interface between the central nervous system and the peripheral blood. The neurovascular unit (NVU) is responsible for the regulation of the greatly discriminating BBB-brain tissue homeostasis in addition to the cerebral blood flow. The NVU is composed of neurons, perivascular astrocytes, microglia, pericytes, endothelial cells (EC), and the basement membrane. These components are connected by close interactions that define them as a single unit and that are challenging to reproduce in vitro (Figure 3). Microfluidic devices have been used to mimic the BBB (Booth and Kim, 2012; Griep et al., 2013) and to model the NVU, combining both a vascular chamber and a brain chamber, allowing for cell-to-cell communication and independent perfusion of both compartments separated by a membrane (Brown et al., 2015). The neurovascular microenvironment has also been mimicked in microfluidic devices using iPSCs (Campisi et al., 2018), recreating in vivo permeability restrictions (Park et al., 2019).

The BBB is disrupted in neurodegenerative diseases and cancer, but the mechanisms are not completely understood, leading to increasing interest in modeling NVU in vitro. Multiple microfluidic-based BBB-on-chip technologies have been described (Oddo et al., 2019), together with platforms that incorporate human tumor and stromal cells in a 3D extracellular matrix, sustained by perfused microvessels (Sobrino et al., 2016). However, current limitations still prevent adequate EC-pericyte and EC-astrocyte interactions, essential for maintaining an appropriate tumor microenvironment (Phan et al., 2017), that strongly influences cancer progression. Cells of the neurovascular unit have an influential role in the outcome of brain metastatic tumor cells (Xu, H. et al., 2016), both in the extravasation through the brain vasculature and the survival mechanisms in the brain parenchyma (Wilhelm et al., 2017). It has been established that instructive factors secreted by endothelial cells can control multiple parameters in brain tumor evolution, including stem cell maintenance and therapeutic response (Ngo and Harley, 2020). It is well accepted that pericytes are essential in the maintenance and integrity of the blood-brain barrier, but just recently the role of pericytes in glioblastoma tumor progression has been reported. The authors show that increase presence of pericytes in GBM blood vessels relates to a poor response to chemotherapy in patients, revealing these cells as potential targets of anticancer therapies (Zhou et al., 2017). These experiments suggest that removal of glioblastoma-derived pericytes alters vascular permeability in the brain tumor, leading to an increased delivery of small molecules. Since glioma stem cells differentiate into pericytes to support vessel structure and support tumor growth (Cheng et al., 2013), the understanding of the pericyte biology in the GBM microenvironment may lead to more efficient therapeutic approaches (Guerra et al., 2018). Recent microvasculature system-on-chip are starting to provide a functional niche to evaluate the dynamics of patient derived tumor cells (Xiao et al., 2019) (Figure 2B).

The rapid growth and invasion, together with a high resistance to available treatments make the prognosis especially poor for patients suffering GBM tumors. A great challenge of brain tumor therapeutics is the efficient delivery into the tumor site, surpassing the exclusive BBB permeability. Even disruptions in the tumor vasculature are not enough to allow effective drug penetration. Microphysiological platforms that recapitulate the specific functionality of the human BBB can provide spatial-temporal monitoring of drug distributions in the vascular and perivascular regions (Ahn et al., 2020). The advances in micro-NVU technology are providing robust platforms for effective preclinical screening of therapeutics for brain diseases (Lee and Leong, 2020). An increasing availability of human glioblastoma samples in these microtissue platforms will be necessary to improve translatability of these data into humans.

3.3. Neural regulation of cancer

Despite the critical role of the nervous system within the brain, just recently the vital contribution of the neurons in cancer growth has been recognized (Jobling et al., 2015). Monje et al. discovered that soluble extracellular NLGN3 protein, involved in synaptic adhesion, was involved in tumor growth, stimulating the PI3K pathway (Venkatesh et al., 2017). These findings opened a new area of cancer treatment strategies. However, few models exist that investigate the cancer cell – neuron interactions, despite the progress in microfluidic devices to explore environmental stimuli on neuroglia, human astrocyte reactivity, and myelin production (Unal et al., 2020; Zambuto et al., 2020). Lei et al. fabricated a compartmentalized microfluidic device that evidenced the support of neurites towards tumor cell migration, that decreased when neuronal signaling was interrupted (Lei et al., 2016) (Figure 2C). Machine learning has been used to investigate different factors that lead metastasis from other organs to the brain (Kingston et al., 2019). These technologies can help develop novel new biomarkers for detecting the early tumor changes that predict brain metastasis (Cho et al., 2021), distinguishing glioblastoma from single brain metastasis (Bae et al., 2020) and providing more affordable diagnoses in brain pathologies (Ker et al., 2019). Although some in vitro models exists (Gregory et al., 2020), there is a lack of models that address nerve-tumor interaction. The application of organ-on-a-chip models to comprehend neural implication in tumorigenesis and cancer progression may also provide a valuable tool for investigating the different factors that attract metastatic cells to the brain.

3.4. Astrocytes

Astrocytes are the most abundant cell population in the brain and perform a variety of functions in health and disease. They contribute to homeostasis by maintaining the blood brain barrier, facilitating ion buffering and neurotransmitter recycling, and signaling the immune system, among other essential roles (Miller, 2018). To accommodate for these diverse roles, astrocytes represent a greatly heterogenous population (Matyash and Kettenmann, 2010). In brain malignancies, tumor-associated astrocytes have been found to contribute to anti-inflammatory responses (Henrik Heiland et al., 2019) and support other functions of tumor survival, invasion and therapeutic resistance (Zhang et al., 2020). Astrocytes undergo diverse phenotypic changes in response to stimuli, altogether called as reactive astrogliosis (Pekny and Pekna, 2014). Due to the high plasticity of these cells, it has been just recently been reported an in vitro platform that provided maintenance of astrocyte quiescence and on demand activation (Galarza et al., 2020). Astrocyte reactivity may lead to a broad variety of changes in astrocyte morphology, molecular expression and functions that can have a great influence in tumor biology. They express damage-associated molecular patterns (DAMP) and pathogen-associated molecular pattern (PAMP) receptors (Sofroniew, 2020).

Astrocytes are a type of central nervous system cell that plays a role in the structure of the brain and in the blood brain barrier (BBB). In the presence of GBM, astrocytes are converted into tumor associated astrocytes. These tumor-associated astrocytes (TAA) then influence the proliferation, migration, survival, and invasion of GBM. TAAs also play a role in the therapeutic resistance of GBM (Brandao et al., 2019). Astrocyte end feet in the brain are displaced by GBM cells allowing for them to invade along blood vessels and diminish the vascular response to the astrocytes (Watkins et al., 2014). Microfluidic devices can be used to further examine this relationship using separate chip compartments to study how secondary brain tumors are formed in different contexts. Compartmentalized microfluidic devices can be used to produce an artificial BBB by culturing astrocytes and endothelial cells, forming a barrier between the brain chamber and the vascular chamber. When cultured alone, aggressive glioma cells were unable to penetrate the BBB while other cancer cell lines were, though when co-cultured with astrocytes only the GBM cells formed a homogenous layer. This findings indicate a possible mode of metastasis and to traverse the BBB (Xu, H. et al., 2016). A microfluidic chip has mimicked lung cancer metastasis to the brain, evaluating changes in cancer cells and detecting damage of astrocytes after cancer cell invasion into the brain (Xu, Z. et al., 2016). Reactive astrocytes have been shown to influence the progression of multiple brain diseases. Their high heterogeneity has led to both positive and negative effects. The understanding of astrocyte biology, especially reactive astrocytes, and the neuro-immune axis, may allow for the improvement of brain cancer treatments.

3.5. Immune system

Some GBM patients show long-term responses to immunotherapies in clinical trials (Reardon et al., 2017). However, the lack of biomarkers to predict patient’s response constitutes an essential element to reveal resistance mechanisms and personalize combinatorial therapies. Biomimetic and photoresponsive hyaluronan (HA) hydrogel nanocomposites have been engineered for the immunomodulation of macrophages (Wang et al., 2019). Recently, microfluidic-based, patient-specific microphysiological systems have been reported that investigate the heterogeneity of the TME and optimize anti-PD 1 immunotherapies based on the different GBM subtypes (Cui et al., 2020). These vascularized 3D hydrogel constructs have also demonstrated the role of macrophage-associated immunosuppression in GBM angiogenesis (Cui et al., 2018) (Figure 2D). The use of these MPS has advantages over previous organotypic models such as the capacity of real-time monitoring of tumor-immune-vascular interactions and therapy responses. The emergence of immune-based therapies has stimulated the development of microfluidic models that are able to quantify temporal levels of tumor-infiltrating lymphocytes (Moore et al., 2018). This system was designed to analyze the interactions between autologous lymphocytes and tumor fragments from specific populations, instead of cancer cells.

Tumor associated macrophages (TAM) represent the largest immune population in GBM tumors, most often constituting up to 30% of the tumor mass, relating also to poor prognosis and therapeutic resistance (Hambardzumyan et al., 2016). It has been shown that GBM tumors secrete immunosuppressive factors, such as transforming growth factor-β1 (TGF-β1), and colony-stimulating factor-1 (CSF-1) that induce immunosuppressive ‘M2-like’ phenotype in monocytes (Thomas et al., 2012). Three dimensional models have shown the possibility of studying GBM-microglia interactions in controlled microenvironments (Chen et al., 2020) and the use of a microdevice platform has enabled a better understanding of factors that affect T cell function for the design of more efficient immunotherapies (Pavesi et al., 2017). Three-dimensional bioprinted brain mimics, consisting of glioblastoma cells and macrophages, have revealed the potential of GBM cells to attract macrophages and the ability of immune cells to support invasion of GBM cells (Heinrich et al., 2019). These mini- brains are presented as tool to study the interactions between these two cell types and to test therapeutics that target those mechanisms.

Increasing evidence suggests that organ preference in metastatic processes is related to the crosstalk between cancer cells and the organ microenvironment. As part of the tissue microenvironment, immune cells can play a key role in the establishment of secondary brain tumors (Boussommier-Calleja et al., 2016). Microphysiological systems detected a bidirectional crosstalk tumor–macrophage, as tumor associated macrophages promoted the invasion of breast cancer cells, while breast cancer incentivized the differentiation of glioma cells into TAM (Mi et al., 2019).

Immune cells are exceptionally diverse, and the resulting cellular interactions lead to a variety of molecular responses. Therefore, there is a need for specialized platforms that allow single cell analyses of immune cells in their distinct microenvironments. MPS provide the desirable characteristics for single-cell monitoring, becoming attractive platforms to understand immune mechanisms (Jammes and Maerkl, 2020). The possibility of measuring single-cell transcriptional profiles related to linage and cell cycle may be particularly valuable in the understanding of brain tumor evolution (Kimmerling et al., 2016). Additionally, the design of micro-devices that provide spatial temporal analyses of single-cell analysis and single-cell RNA-seq profiling, in combination with bioinformatics resources (Huang et al., 2009), will enable new strategies to understand the heterogeneous responses of immune system at the single-cell level (Huang et al., 2020; Wimmers et al., 2018).

4. Conclusion and future considerations

The human pathophysiology of neural diseases is highly complex and affects the development of preclinical models. Due to the difficult access to human brain, tumor chips hold a great potential to revolutionize cancer treatments by offering a higher prediction power of clinical trials than current models. These tumor-on-a-chip contribute to more personalized analysis, recreating patient-specific tumor microenvironmental characteristics. In spite of this potential, scientific community is still facing several challenges, such as the isolation and maintenance of the individual’s diseased cells (stromal, cancer and immune cells) in conditions that ensure their original phenotypes, and rapid and efficient integration into the microfluidic devices. Additionally, alternatives to PDMS are necessary to fabricate the devices, materials flexible and optically clear but that avoid non-specific absorbance of biologically active molecules. In this area, progress in bioprinting is allowing the construction of increasingly complex perfused platforms.

The great advantage of tumor chips lies in the possibility of studying specific microenvironmental components on cancer cells. This technology allows for the addition of multiple cells and biomechanical cues, obtaining data in real time with high resolution. Additionally, the possibility of connecting multiple organ chips will provide more information about metastatic brain tumors. The advances in induced pluripotent stem cells isolated from patients increase the translational potential for more personalized brain tissue devices, that also allow for multi-omics analyses. The combination of miniaturized organotypic platforms with microfluidic devices have uniquely enabled the discovery of new cell-cell and cell-ECM interaction that affect tumor growth an invasion. The development of these systems requires the meticulous adjustment of cell culture medium composition to each cell type, adapted to the unique characteristics of the tissue. In addition, future refinement of these platforms will need the use of accurate compositions to obtain more standardized formulations. Questions still exist around the need of prolonged culture viability combined with real-time, label free read-out technologies. Challenges in biofabrication technologies include the immaturity of the systems, the streamlining of the device processes, validation and increased reproducibility. Nonetheless, the fabrication costs are decreasing, and significant efforts are focusing on the improvement of cell differentiation control. Moreover, the combination with bioelectronics will also improve the diagnostic and prognosis capabilities.

Some technical challenges in cell manipulation, monitoring and readout performance still need to be overcome for these systems to be established as alternatives to animal models. However, the organ-on-a-chip technology provides a cost-effective and fast platform that can allow a groundbreaking progress in our understanding of tumor microenvironment, how these ecosystems respond to alterations and the interaction with other organs. These systems’ applications surpass the drug screening and have the capacity of providing a more accurate evaluation of the complex interactions between the tumors and the environmental stimuli.