The Advances in Glioblastoma On-a-Chip for Therapy Approaches

Simple Summary This systematic review showed different therapeutic approaches to glioblastoma on-a-chip with varying levels of complexity, answering, from the simplest question to the most sophisticated questions, in a biological system integrated in an efficient way. With advances in manufacturing protocols, soft lithography in PDMS material was the most used in the studies, applying different strategy geometrics in device construction. The microenvironment showed the relevant elaborations in co-culture between mainly human tumor cells and support cells involved in the collagen type I matrix; remaining an adequate way to assess the therapeutic approach. The most complex devices showed efficient intersection between different systems, allowing in vitro studies with major human genetic similarity, reproducibility, and low cost, on a highly customizable platform. Abstract This systematic review aimed to verify the use of microfluidic devices in the process of implementing and evaluating the effectiveness of therapeutic approaches in glioblastoma on-a-chip, providing a broad view of advances to date in the use of this technology and their perspectives. We searched studies with the variations of the keywords “Glioblastoma”, “microfluidic devices”, “organ-on-a-chip” and “therapy” of the last ten years in PubMed and Scopus databases. Of 446 articles identified, only 22 articles were selected for analysis according to the inclusion and exclusion criteria. The microfluidic devices were mainly produced by soft lithography technology, using the PDMS material (72%). In the microenvironment, the main extracellular matrix used was collagen type I. Most studies used U87-MG glioblastoma cells from humans and 31.8% were co-cultivated with HUVEC, hCMEC/D3, and astrocytes. Chemotherapy was the majority of therapeutic approaches, assessing mainly the cellular viability and proliferation. Furthermore, some alternative therapies were reported in a few studies (22.6%). This study identified a diversity of glioblastoma on-a-chip to assess therapeutic approaches, often using intermediate levels of complexity. The most advanced level implemented the intersection between different biological systems (liver–brain or intestine–liver–brain), BBB model, allowing in vitro studies with greater human genetic similarity, reproducibility, and low cost, in a highly customizable platform.


Introduction
Glioblastoma (GBM) is the most common primary malignant brain tumor in adults. The annual incidence of GBM in the United States is 3.23 cases per 100,000 people [1] and is one of the most fatal malignant diseases in humans. The patient median survival is

Inclusion Criteria
We included only original articles published in English in the last ten years, with the available full text, and that used microfluidic devices to evaluate different therapeutic approaches for glioblastoma tumor models developed from the culture of tumor cells.

Exclusion Criteria
We excluded articles that did not report the therapeutic approach in the microfluidic device, that did not perform a tumor microenvironment reconstitution from the use of tumor cells, as well as articles indexed in more than one database (duplicates), review articles, letters, articles in press, communications, book chapters, abstracts, incomplete articles, editorials, and expert opinions.

Data Extraction
In this systematic review, the collected data were segregated into the following topics: (i) the microfluidic device design, their material used, and its manufacturing method; (ii) the characteristics of the cells used in 3D culture and the medium; (iii) the microenvironment reconstitution for the glioblastoma model and their maintenance; and (iv) the therapeutic approaches applied in the devices and the techniques used for the therapeutic efficacy evaluation.

Data Analysis
The percentage distribution, obtained for each variable analyzed in the tables was used to characterize and present all of the results. Each study was classified into 3 categories of complexity, from (+) to (+++), based on how each topic was approached separately in each table. Finally, we considered the analysis of the results reported in the tables and applied a generic classification of device complexities in four categories (I-IV).

Overview of the Reviewed Literature
We searched publications of the last 10 years, considering the period between September 2011 and September 2021, indexed in PubMed and Scopus, and a total of 446 articles were identified. Of the 119 articles found in PubMed, 94 were excluded after screening (89 duplicated in Scopus search, and 5 reviews), and 22 articles were excluded after eligibility analysis (12 articles did not report the therapeutic approach used for glioblastoma on-a-chip, 6 articles reported only the usage of the microfluidic device for analysis of the part of the experiment, such as CHIP-Seq or CHIP-qPCR, and 4 articles developed the study in silico), thus, only 3 articles were included from this database. Of the 327 articles identified in Scopus, after screening, 48 articles were excluded (18 reviews, 15 conference papers, 8 book chapter/series, 2 notes, 2 publications in other languages, 1 conference review, 1 editorial, and 1 short survey), and 260 articles were excluded after eligibility analysis (120 articles did not report the therapeutic approach used for glioblastoma on-a-chip, 110 articles reported only the usage of the microfluidic device for analysis of part of the experiment, such as CHIP-Seq, CHIP-qPCR, and chromatin immunoprecipitation-CHIP, and 30 articles developed the study in silico), thus, only 19 articles were included from this database. As shown in Figure 1, only 22 unduplicated full-text articles were included in this review , and the histogram and spider chart show the distribution of articles by year and research centers, respectively.

Methods Cultivation of Cells Used in the 3D Culture
Regarding the 3D culture of glioblastoma model in a microfluidic device (Table 3), the culture methodology that involves the ECM components (concentration and volume) and cell types (concentration, culture time, medium change, and their flow rate) that represent important aspects for the development of a tumor model biomimetic to evaluate the different methodologies and therapeutic agents, was analyzed. Figure 2. The schematic figures of glioblastoma on-a-chip devices for therapy approach used in some of the selected studies of this systematic review. (A) The integrated microfluidic system for single-cell separation and sphere formation, adapted with permission from [46], the American Chemical Society. (B) 3D co-culture unit generative process and the analysis of the confocal images of the chip, showing the HUVEC cells in the lumen, adapted with permission from [52], Analytica Chimica Acta. (C) MCF7 and U87MG cancer cells diagonally seeded into square-shaped microchambers, in the hydrogel microfluidic device, and analysis of confocal microscopy images, adapted with permission from [53], Electrophoresis. (D) Magnetohyperthermia process in tumor-on-a-chip using magnetic nanoparticles dispersed in aqueous medium submitted to an alternating magnetic field., adapted with permission from [5], Einstein. (E) A microfluidic platform mimics the blood-brain barrier (BBB) using two PDMS sheets a polycarbonate membrane. BBB unit was directly connected to the µSPE unit for mass spectrometry detection., adapted with permission from [56], Analytica Chimica Acta. (F) Biomimetic design of miniaturized artificial perivascular niche on a chip for analysis on chemoresistance in GSCs and endothelial cocultured and relative metabolites by liquid chromatography mass spectrometry, adapted with permission from [49], Analytical Chemistry. (G) The closed-loop acoustofluidic device with multilayer for drug release in a tumor by the focal ultrasound system, adapted with permission from [55], Small. (H) Glioblastoma on-a-chip comprised of tumor and tumor-associated stroma compartments with side channels (delivered nutrients and drugs), and the actual image of the fabricated model., adapted with permission from [42], International Journal of Molecular Sciences. (I) Simplified photodynamic therapy of methylene blue conjugated polyacrylamide nanoparticles, with a polyethylene glycol dimethacrylate cross-linker on microfluidic chip, adapted with permission from [59], Chemistry of Materials.    The GBM cells were seeded into both inlet channels simultaneously. The cells were captured in the microwells and cultured for 7 days A half medium 48-72/NR +  Another relevant aspect in 3D co-culture is the ECM addition, which was reported in 73% of the selected studies [39,[42][43][44][45][47][48][49][50][52][53][54][55][56]58,59]. Collagen type I represented 35%, being the most used in ECM composition [39,[42][43][44][45]47,48,55,58,59], followed by 15% Matrigel [43,50,54,56], and a smaller proportion (4%) BdECM [44], fibronectin [49], TGgelatin [52], GelMA [53], and agarose [56]. Of these studies, 45.8% reported ECM concentration used ranging from 0.1 to 12 mg/mL [39,[42][43][44][45]48,50,[53][54][55][56], and only 16% reported the volume administration, ranging from 8 to 20 µL [43,47,48,53]. In contrast, 20% did not use any ECM components [40,41,46,51,60]. The culture methodology in the microfluidic device focused mainly on the order and position of cell culture, from the treatment of the device with ECM to maintenance after the culture. Some strategies were used to promote the formation of a 3D matrix, inverting the device surface [45,48].
For  [39,54] and adenocarcinoma [54]) were used. Furthermore, 31.8% of studies used some supporting endothelial cells, with one or more of these cells combined (astrocytes, hCMEC/D3, HUVEC, HBMEC, BMEC, Eahy926, and Bend3), being prevalent in the co-culture with the first three of these cells (20% each), aiding in chemical communication and secretion of the ECM, obtaining results closer to those obtained in vivo experiments [61][62][63][64]. In relation to the cell type, we also analyzed the concentration used, which varied between the different types, as well as within the same type, for example, the U87 number cells ranged from 10 4 to 10 7 cells/mL. Cell culture time of 92% of the studies was from 0.125 to 10 days and medium change during culture was reported in 54% of the studies, being carried out from 2 to 72 h. The flow rate was reported in only 37.5% of the studies, ranging from 0.5 to 4.7 × 10 3 µL/min, and 8.3% of studies did not apply the shear rate [56,59], an important factor for tumor growth.
We established a global classification of the glioblastoma on-a-chip model for therapeutic approaches, at different levels of complexity (I-IV), with level IV being the most complex, based on all aspects investigated in the present study and the results presented in the tables. The studies were classified considering their design and fabrication; cell culture isolated or co-cultures, ECM complexity, besides the therapeutic approaches used. This way, few studies (4.5%) were classified with low level of complexity due to their used simple shape, a single-cell type culture, without ECM, and a simple therapeutic approach [42,52]. Levels II and III often already had the most complexity reported with 36.4% [39,40,43,49,58] and 40.9%, respectively [45,47,48,50,51,[53][54][55]57,59,60], shown to improve the design complexity through the use of the concentration gradient, as also parallel chambers with interconnections through pores, or the use of some type of ECM. Level III was regarded as the use of co-culture, advanced therapeutic approach, or the improvement of criteria used in level II. Of the studies, 18.2% were classified as level IV due to the use of intersection between different biological systems (liver-brain or intestine-liver-brain), BBB model, tri-culture, ECM adaptation, or use the synthetic scaffold [41,44,46,56].
In terms of device material, PDMS was the most used (72%) and it has a variety of advantages, including being durable, inert to most materials (patterned or molded), and chemically resistant to many solvents. However, this material also suffers from high compressibility, which causes a seal's shallow relief features to deform, bend, or collapse. The molding step is facilitated by the elasticity and low surface energy of the PDMS, which also gives the possibility to replicate the size and shape of the features present in the mold by mechanical deformation. In addition, PDMS molds can be manufactured from a single master [66]. Among substrates, glass was most often in the reviews (64%) [41][42][43][44][45]48,[50][51][52][53]55,[58][59][60], followed by PDMS (18%) [40,46,49,54], nano/microfluidic glass channels giving improved control of the chemistry in the microsystem; PDMS or other polymers are already often used due to their low-cost fabrication process [67], but they are chemically active and strongly absorb proteins to their surface unlike glass channels, which are inert to most chemicals. Furthermore, glass channels are easy to clean, maintain, reuse, and very efficient in microscopic analysis due to optical characteristics [68].
For the glioblastoma model, the most commonly used cells include human-derived cell lines, such as U87 and U251, and mouse cell lines C6 and F98. U87 human GBM was the most reported in the selected studies (25.6%) [39][40][41][42][44][45][46][48][49][50][51][52][53][54][56][57][58] as an alternative preclinical testing model, following by the use of U251 [42,46,54,56] and C6 cells [43,47,59,60] (10.3% each). All these cell lines exhibit similar morphological characteristics regarding GBM nuclear pleomorphism and high mitotic index, except F98, which resemble anaplastic glioma. The most aggressive and invasive model is the F98, while the C6 has moderate invasiveness. U87 exhibits profuse neovascularization and has been used to study angiogenesis. When comparing U87 to U251, it was observed that the U87 cells exhibited a significantly higher rate in relation to their proliferation, invasion, and migration [69], and this difference was also observed in the 3D model, showing a rapid migration, and the highest invasion ability (the length of protrusions and the number of cells, invading into the collagen) [70]. For microenvironment studies, C6 has been well used because it resembles human GBM immune infiltrates, being considered a good model of an immunocompetent host for in vivo studies, due to its ability to cause a moderate immune response, as well as U87 and U251 cells. Other tumor cells, such as HepG2 cells, a liver hepatocellular carcinoma, were also used in the same device to compare the therapeutic efficacy and metabolization in different types of drugs and the interaction of the brain-liver system [39]. Another study also used these cells associated with Caco-2, human colorectal adenocarcinoma cells, to evaluate the intestine-liver-glioblastoma biomimetic system [54]. Already, MCF7 human breast cancer cells were used only in regard to glioblastoma therapeutic efficacy and cell migration [53].
The ECM is another relevant aspect for microenvironment formation inside the microfluidic device, since helps cells to attach, communicate, and provides physical scaffolding to biochemical and biomechanical processes, necessary for tissue morphogenesis, differentiation, homeostasis, and other cell functions [71,72]. Currently, a wide range from natural proteins to synthetic scaffolds has been used for culturing cells in a 3D environment. The choice of a suitable matrix depends on the cell type being used and specific experimental objectives. Materials of natural origin are commonly used, to mimic several key features of the native ECM, such as type I collagen, hyaluronic acid, laminin, fibronectin, gelatin, alginate, as well as Matrigel (ECM extracts) [73]. Another potential alternative to Matrigel is GelMA, a natural ECM reported in this review [53].
Of selected studies, 73% used some type of ECM, and collagen type I was the most reported (35%) [39,[42][43][44][45]47,48,55,58,59], as well as being the most important ECM component with which cancer cells interact during their growth. It is the preferred substrate for the adhesion and migration of these cells and also stimulates their invasive behavior. This provides a strong rationale for the use of collagen I matrices in investigations pertaining to invasive behavior by cancers and metastasis [74]. The second component most used in the review was Matrigel (15%) [43,50,54,56], this material is chemically similar to the major components of basement membranes, imparting strength and integrity, but is unable to mimic the barrier function of intact basement membranes due to lesser resistance to cell penetration [74]. This material showed an influence on the spheroids' formation due to its composition that contains a high percentage of laminin, collagen IV, enactin, proteoglycans, and growth factors, associated or not with other natural polymers (collagen, chitosan, hyaluronic acid) or synthesized (polyethylene glycol). Both materials can be used in different applications for generation of 3D models (organoids, primary tissue culture, embryonic stem cells, or induced pluripotent stem cells), tissue explants, and cellular differentiation, suggesting their importance in the composition and function of the matrix. Furthermore, the main components of the matrices cited above can influence the metabolism of cancer cells, as well as interfere with cell signaling and tumorigenesis [74,75].
In addition to ECM, some studies reported the use together with support cells as the astrocytes, HUVEC, and hCMEC/D3, for the formation of the 3D model in the microfluidic device. The supporting cells are able to modulate and produce ECM through secreted factors [76,77]. HUVEC was also used to evaluate the ability of angiogenesis [78,79]. Only the study by Tricinci [41] did not report the use of some type of ECM, substituted by the use of a synthetic scaffold as a similar function.
The HUVEC co-cultures with human glioma cells (U87-MG and T98) resulted in vascular sprout formation. However, no vascular sprout formation was observed in HUVEC co-cultured with human teratocarcinoma cells (NT2), which do not produce VEGF, suggesting their importance in the angiogenesis role. Hypoxia condition in gliomas is another way to lead to the upregulation of VEGF expression and angiogenesis, enhancing tumor growth through neovascularization [80]. EA.hy926 is another endothelial cell used in the neoangiogenesis model and has the advantage of being reproducible, not depending on the primary tissue nor having differences in response along with the genetic variation of each sample as HUVEC [81].
Some microfluidic devices in this review [39,40,56] are aimed at the blood-brain tumor barrier (BBTB) study in the microenvironment, since the molecular selectivity of BBB allow homeostasis in physiological conditions, as also shields the neoplastic cells by blocking the delivery of peripherally administered chemotherapies. For this BBTB, different types of support cells were used, such as hCMEC/D3, when in co-culture with astrocytes has been reported to restore some of the BBB-differentiated phenotypes of isolated brain endothelial cells (BECs) by having a particular impact on the expression and maintenance of tight junction (TJ) proteins [82]. However, the systematic review of human BBTB models of brain permeability for novel therapeutics [82] showed a great variability of cell origin (stem cell-derived, primary or immortalized), monoculture versus co-culture, and other parameters that affected model success. These aspects may cause the under or overestimation of drug permeability and therapeutic efficacy. Nevertheless, also highlighted by some important analyses, as the co-cultures of ECs with astrocytes, and pericytes had significantly upregulated protein or mRNA expression of tight and adherent proteins, and transporters, irrespective of whether pluripotent stem cell, primary, or immortalized cells lines were induced. The murine brain microvascular endothelial cells (bEND3), also used in this review [55], are known to be successful in forming barriers when co-cultured with astrocytes [83]. Astrocytes, in turn, have widely distinct morphological, molecular, and functional properties, suggesting the existence of heterogeneous subpopulations and have been identified as modulators of the BBB permeability, as well as their impact on TEER, and gene expression [84].
We also analyzed the methodological steps used in the formation of microenvironments within the device, being relevant to note that 77.3% reported details of the construction of this microenvironment, in which 31.8% of studies firstly covered the device with some type of ECM followed by cell culture, in a staggered isolated or in co-culture manner. At the same frequency, the simultaneous infusion of matrix and cells was observed. Only two studies (9.1%) reported the inversion of the PDMS surface to form the 3D structure [45,48], and the study by Qu et al. [45] reported the use of type I collagen after the formation of spheroids. In parallel, the flow of the medium responsible for the shear rate was described in only eight studies (36.3%), which is a relevant aspect regarding the cell phenotype, as well as for the renewal of the medium and nutrients.
New combined therapeutic approaches that involved using advanced technology were applied in this review, such as 9.1% of studies that used nanotechnology resources and focused ultrasound to improve the drug delivery in chemotherapy [41,55]. In addition, alternative approaches were used as photodynamic therapy (13.6%) [53,59,60], and 4.5% of them use gold and iron nanoparticles associated with near-infrared laser [53], and alternating magnetic field [43], respectively, that promote the death of tumor cells by hyperthermia. Despite the low frequency of studies with these alternative therapies, scientific interest in this area has grown due to technological advances and the development of multifunctional probes capable of being applied in translational studies, combining more than one therapy and analysis technology. The 3D model allows the application of these technological advances in models that use mainly human tumor cells, capable of predicting more mimetic responses than in vitro and in vivo studies.
Therefore, considering all aspects involved in this review, we used the classification system in four levels of glioblastoma on-a-chip fabrication complexity. The increased complexity in the elaboration of the devices verified in this review (18.2% level IV [41,44,46,56], 40.9% level III [45,47,48,50,51,[53][54][55]57,59,60], and 36.4% level II [39,40,43,49,58]) has reflected the diversity of components present in the real tumor microenvironment, as well as the responses obtained from the interaction of the different systems, cells and ECM, as described in the study by Jo et al. [50] that observed chemoresistance regarding the use of Matrigel in the treatment with DOX, among other aspects. In this sense, the literature has proven that the responses of organ-on-a-chip studies have been increasingly closer to in vivo studies than the results obtained in in vitro [86,87], and represents an excellent platform for validation of therapeutic processes for glioblastoma tumors.
This review showed the current aspects of glioblastoma research through the organ-ona-chip device for therapeutic approaches, but the diversity features of device elaboration and approaches did not allow us to conclude which was the most effective therapeutic approach among the studies, being a relevant limitation of this study.

Conclusions
This systematic review identified the diversity of glioblastoma on-a-chip to assess therapeutic approaches, with different levels of complexity. We found that soft lithography, a printing process with a high micrometric resolution, and PDMS, a biocompatible and chemically resistant substance, were found to be the most used in this review. The tumor microenvironment was mainly composed of ECM rich in collagen type I associated with human tumor cells, cultivated in a 3D framework. Chemotherapy remains a more studied approach, alone or in comparison with other therapeutic alternatives, in the search for more efficient ways of drug delivery, with fewer collateral effects, using nanocarriers associated with drug activation techniques and hyperthermia promotion for treatment of glioblastoma.
In terms of experiment complexity, a few researches have shown a low level of complexity by using a simple shape, unicellular culture, without ECM, and a straightforward treatment strategy. The adoption of a concentration gradient, parallel chambers with interconnections, or some sort of ECM were among the most often reported intermediate complexity (level II and III) features. The usage of co-culture, an advanced therapeutic strategy, or an enhancement of the criteria employed in level II, were all explored in the third level. Interestingly, the most advanced level implemented the intersection of different biological systems (liver-brain or intestine-liver-brain), the BBB model, tri-culture, ECM adaptation, or the use of synthetic scaffolds, allowing for the recognition of advances in this technology for organ-on-a-chip studies with greater human genetic similarity, reproducibility, and low cost on a highly customizable platform. Thus, finally, we can conclude that, taking into account the studies included in the review, the glioblastoma-on-a-chip platform is an excellent alternative for evaluating the therapeutic process of this type of tumor with high reproducibility.