Effectiveness of Bioinks and the Clinical Value of 3D Bioprinted Glioblastoma Models: A Systematic Review

Simple Summary Glioblastoma is the most malignant cancer of the glioma series, and it is highly invasive. The progression and recurrence of glioblastoma remain common due to the development of drug resistance. Of the current disease models and strategies used in pre-clinical studies for drug testing, three-dimensional (3D) bioprinting is an emerging technology in constructing a glioblastoma model. In this paper, 19 out of 304 articles yielded from the database search were selected and analysed through a systematic process. The selected studies present the effectiveness of different bioinks, which were used to mimic the tumour microenvironment of glioblastoma in bioprinting. The clinical value of the 3D bioprinted glioblastoma models on the efficacy of treatments or drug response was evaluated. Abstract Many medical applications have arisen from the technological advancement of three-dimensional (3D) bioprinting, including the printing of cancer models for better therapeutic practice whilst imitating the human system more accurately than animal and conventional in vitro systems. The objective of this systematic review is to comprehensively summarise information from existing studies on the effectiveness of bioinks in mimicking the tumour microenvironment of glioblastoma and their clinical value. Based on predetermined eligibility criteria, relevant studies were identified from PubMed, Medline Ovid, Web of Science, Scopus, and ScienceDirect databases. Nineteen articles fulfilled the inclusion criteria and were included in this study. Alginate hydrogels were the most widely used bioinks in bioprinting. The majority of research found that alginate bioinks had excellent biocompatibility and maintained high cell viability. Advanced structural design, as well as the use of multicomponent bioinks, recapitulated the native in vivo morphology more closely and resulted in bioprinted glioblastoma models with higher drug resistance. In addition, 3D cell cultures were superior to monolayer or two-dimensional (2D) cell cultures for the simulation of an optimal tumour microenvironment. To more precisely mimic the heterogenous niche of tumours, future research should focus on bioprinting multicellular and multicomponent tumour models that are suitable for drug screening.


Introduction
Glioblastoma (World Health Organisation (WHO) grade IV glioma) is the most common primary brain cancer. It is highly invasive and the most malignant of the glioma series of cancers [1,2]. Glioblastoma patients have a poor prognosis, with a median survival time

Search Strategy
The review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [13]. A comprehensive literature search was performed on five electronic databases, namely PubMed, Medline Ovid, Web of Science, Scopus and ScienceDirect. The keywords were ("3D bioprinting" OR "3D printed" OR "three-dimensional in vitro model") AND ("glioblastoma" OR "glioblastoma multiforme"). This systematic review was registered in PROSPERO (313977).

Study Selection and Eligibility Criteria
The titles and abstracts of all the identified records were screened independently by two authors (S.W.L and S.C.T) to select potentially relevant studies based on pre-designated inclusion and exclusion criteria. Any disagreement was resolved by a third author (S.Y.L). The inclusion criteria were as follows: (1) used bioinks, (2) in vitro and in vivo studies, (3) type of glioblastoma cell lines or cells derived from glioblastoma and the investigations performed, (4) used 3D bioprinted scaffolds, and (5) original article written in the English language only. The exclusion criteria were as follows: (1) non-particular interest in 3D bioprinting, (2) systematic and narrative reviews, interpretations, case series, guidelines and technical reports.

Data Extraction
The full texts of the articles that met eligibility criteria were further reviewed, and data extraction was independently performed by two authors (S.W.L and S.C.T). The following data were extracted from the included studies: (1) study information (authors, year of publication and study design); (2) intervention details (biomaterials and cells used, crosslinking methods and materials, 3D bioprinting techniques and drug testing); (3) outcome details (rheological and morphological characteristics of bioinks, biological characteristics of glioblastoma cells such as cell viability or cell proliferation or cell migration, and drug response). Disagreements were resolved by discussion amongst the authors (S.W.L and S.C.T), with the advice of a third author (S.Y.L) when necessary.

Quality Assessment
The quality of the included studies was appraised independently by the same two authors (S.W.L and S.C.T) using the proposed checklist by the Joanna Briggs Institute (JBI) [14]. JBI is an international research organisation established at the University of Adelaide's Faculty of Health and Medical Sciences in Australia to support evidence-based healthcare and research. The JBI critical appraisal tool was used to assess a study's methodological quality and identify the possibility of bias in its design, conduct, and analysis [14]. Each item on the risk of bias tools was scored with A (indicating low risk of bias), B (indicating high risk of bias), C (indicating bias not clear) or D (indicating not applicable).

Quality Evaluation
The risk of bias in the included studies was assessed using a checklist In general, almost all studies had a low risk of bias. There were, howev [23,26] that did not explicitly state whether there was a control group. Th risk assessment are summarised in Table 1.

Quality Evaluation
The risk of bias in the included studies was assessed using a checklist of the JBI [14]. In general, almost all studies had a low risk of bias. There were, however, two studies [23,26] that did not explicitly state whether there was a control group. One [27] study was suspected high risk of bias for the multiple outcome assessments taken before and after the intervention/exposure. Three studies [8,17,20] were not clear with reliable outcomes measured or without number of replicate experiment, and one study [20] was found unclear without method of statistical analysis. Follow-up was not applicable to all the included studies. The results of the risk assessment are summarised in Table 1. Wang et al. [15] Wang et al. [16] Dai et al. [3] Han et al. [17] Haring et al. [18] Heinrich et al. [19] Tricinci et al. [20] Utama et al. [21] Yi et al. [22] Lee et al. [8] Wang et al. [23] Wang et al. [24] Chaicharoenau-domrung et al. [25] Ba-kirci et al. [26] Smits et al. [27] Tang et al. [28] Chadwick et al. [29] Hermida et al. [30] Tang et al. [31] Checklist No confusion about which variable comes first The subjects involved in any of the comparisons were comparable Other than the exposure or intervention of interest, the subjects involved in any comparisons received similar treatment/care There was a control group Multiple outcome assessments taken before and after the intervention/exposure Participants' results measured in the same way in any comparisons

Cell and Animal Models
The human glioblastoma cell line U87 was used in the majority of in vitro research (47.4% of the included studies). Two studies used only human glioblastoma cell line U87 [8,26]; one study used human glioblastoma cell line U87 and glioma stem cell line SU3 [3]; one study co-cultured human glioblastoma cell line U87 with human astrocytes [27]; one study co-cultured glioblastoma cells U87 with human vascular endothelial cells (HU-VECs) and lung fibroblasts (LFs) [17]; one study co-cultured human glioblastoma cell line U87 and glioblastoma stem cell lines (G7, G144, G166) with monocytic MM6 [30]; one study co-cultured normal U87 cells and GFP-expressing U87 with human cerebral microvascular endothelial cell line (hCMEC/D3) [20]; and one study involved human glioblastoma cell line U87vIII and neuroblastoma SK-N-BE(2) [21]. In two studies, the human glioma cell line U118 was co-cultured with the human glioma stem cell GSC23 [16,24]. In five studies, only one cell line was used, namely the human glioma cell line U118 [15], human glioma stem cell GSC23 [23], human glioblastoma cell U-251 [25], human glioblastoma cells D54-MG [18] and GL261 mouse glioblastoma cell line [19], respectively. In addition, a total of four studies utilised primary cells from the patients. For example, one study used patientderived glioblastoma cells alone [29]; one study co-cultured human glioblastoma cell line U87 cells and patient-derived glioblastoma with HUVECs [22]; one study co-cultured human patient-derived glioblastoma stem cells (GSCs) TS576 with HUVECs [28]; and one study co-cultured patient-derived GSCs with macrophages, astrocytes and neural stem cells (NSCs) [31]. Additionally, there were three in vivo studies of 3D bioprinting which utilised 4-6 week old nude mice; however, the studies did not disclose the number of animals used [15,16,31].

Physical Properties and Biocompatibility Measures
The properties of bioinks and their impact on cell morphology, biological characteristics and drug response are shown in Table 3. The included studies reported the pore size of hydrogels, which varied from 2-400 µm, while the porous percentage ranged from 53-89%. The alginate bioink with pore sizes of 100-400 µm and a porosity of 89% was found to be able to preserve cell viability at around 78% even after being bioprinted for 21 days, whereas the GAF bioink with pore sizes of 338.41 ± 23.18 µm was found to be able to preserve cell viability at around 89% even after being cultured for 15 days [3,19,23,25,31].

Overview of the Included Studies
This systematic review of 19 studies showed that bioinks were effective in simulating the glioblastoma tumour microenvironment, and the models can contribute to a more accurate drug testing. Most 3D bioprinted cell culture models demonstrated excellent cell viability and cell proliferation within the scaffolds. Moreover, 3D bioprinted glioblastoma models showed higher resistance to drugs when compared with the conventional 2D cell cultures, indicating that the 3D bioprinted models represented the in vivo morphology and complex tumour microenvironment better than 2D cell cultures. In a 3D environment, co-culturing of a tumour with macrophages, astrocytes or HUVECs also resulted in higher cellular viability and drug resistance as compared to mono-culturing of a tumour.
3D printing is an emerging technology in brain cancer modelling and drug screening. In spite of increasing published literature that suggested the advantages of 3D bioprinting, most of the studies were pre-clinical, with small sample sizes. In addition, there is a lack of a comprehensive systematic review of the context of bioink materials, material characteristics and effects, and bioprinting strategy. To our knowledge, this is the first systematic review to assess the effectiveness of bioinks in the 3D bioprinting of glioblastoma models and their clinical value.

Bioink Materials and Combination
Alginate has been extensively utilised as a bioink with or without a combination with other matrix components [32]. Wang et al. reported that the combination of alginate and gelatin produced good shear-thinning properties, sufficient mechanical strength after crosslinking and excellent physicochemical properties [16]. The bioink scaffold had welldefined pores for nutrient and oxygen exchange, subsequently supporting the cell viability.
Cells have receptors but not for alginate. The alginate was chemically modified using carbodiimide conjugated to an Arg-Gly-Asp peptide sequence (RGD) to improve cell-cell interaction and matrix interaction [33,34]. When compared to unmodified alginate, RGDalginate showed an expansion of cells and the formation of cell aggregate in less than 24 h. U87MG cells were able to retain high viability in RGD-alginate matrices and grew in culture for more than a month [30]. In addition, cell proliferation increased when the density of RGD increased [35,36].
Alginate has also been used in combination with fibrin and genipin [8,27], and the addition of fibrin in the hydrogels was found to enhance the expansion of stem cells and tumourigenic cells [37][38][39]. The polymerisation of a fibrin-based bioink with a combination of alginate and genipin was initiated by the enzyme thrombin. A chemical crosslinker, e.g., calcium chloride (CaCl 2 ), is commonly used to crosslink alginate, which interacts with the calcium ion (Ca 2+ ) binding sites of fibrin, while genipin interacts with the amine groups to enhance polymerisation and the degree of lateral aggregation and enables stabilisation of the fibrin network [40]. Genipin has also been demonstrated to crosslink fibrin [39] and chitosan [41], successfully slowing down scaffold degradation [8]. In addition, the combination of gelatin with alginate and fibrinogen (GAF) was introduced due to its contribution to high structural stability and suitability for long-term cell culture. Gelatin was selected because it featured cell-binding motifs and allowed physical crosslinking at low temperatures [17].
The ECM is made up of a variety of macromolecules that dictate the tissue's unique biochemical and biomechanical properties, and it plays an important role in cancer progression. As it is difficult to optimally reproduce the intrinsic complexity of native ECM, decellularised ECM (dECM) is one of the options for producing a microenvironment similar to that of the parental tissue. Cancer cells seeded onto dECM have recently been shown to have elevated expression of genes associated with invasion and enhanced interactions between the cells and ECM molecules [42,43]. Porcine brain was successfully decellularised and formed a printable dECM bioink with a serial treatment of chemical and enzymatic agents. Glioblastoma cells showed a higher proliferation rate and enhanced invasion capability in the brain dECM gel compared to the glioblastoma cells in collagen gel. In addition, HUVECs in the brain dECM gel showed an increased expression of the genes related to cell junction molecules and ECM remodelling protein (matrix metalloproteinase 9). As such, tubule networks were formed more actively in the brain dECM gel than in the collagen gel over two weeks [22].
In healthy brain tissues, HA is the most abundant ECM component. HA promotes glioblastoma migration by regulating glioblastoma invasion via the receptor for hyaluronanmediated motility (RHAMM) and CD44, as well as other mechanical and topographical signals [44]. Recent studies showed that a combination of 0.25% HA with gelatin methacrylate (GelMA) had successfully enhanced glioblastoma stem cells' pluripotency and resistance [28,31]. The results were in line with the findings reported by Pedron et al. [45]. GelMA also served as a stiffness modulator, resulting in acceptable mechanical qualities with minimal biochemical cues [31].

Physical Properties of the Bioink Scaffolds
In bioprinting, the architectural design of a tumour model is critical. An interconnected porous structure of bioink is beneficial to mimic the vasculature for nutrient and gaseous exchange, allowing surrounding cells to maintain high viability and functionality. Chen et al. discovered that a 3D microenvironment generated by 3D porous scaffolds not only promoted the formation of tumour cell spheroids but also greatly increased the invasiveness and chemotherapeutic resistance of tumour cells cultivated on 3D scaffolds compared to 2D models [46]. Moreover, Druecke et al. found that the pore size of the scaffold is a key factor of scaffold vascularisation as they discovered that blood vessel ingrowth was significantly accelerated in those scaffold pores with a size larger than 250 µm compared to those with a size less than 250 µm [47]. Hence, optimal pore size is deemed crucial in the fabrication of a functional 3D cell culture system for effective glioblastoma disease modelling.
Shear-thinning is an important parameter in 3D bioprinting because a bioink with excellent shear-thinning quality can minimise clogging during the printing process and immediately restore the structural consistency after printing so that the next layer can be supported [48][49][50]. The shear force generated during printing can be reduced by using a suitable nozzle diameter and low-viscosity hydrogels [51]. For example, to reduce the shear force, cells were integrated with 10% gelatin bioink using a nozzle with a diameter of 0.26 mm at a controlled temperature of 25 • C, and the chamber temperature was lowered to 10 • C during printing to increase the shape fidelity of printed scaffolds [23].
Stiffness is another feature that indicates the ability of bioink scaffolds to withstand mechanical force. This parameter could be quantified by graphing the stress-strain curves of the scaffolds under mechanical pressure and computing the slope of the curve, also known as Young's modulus [52]. The Young's modulus is a critical parameter of biomaterials that influences cell proliferation and differentiation direction [53]. In a study performed by Chaicharoenaudomrung et al. the Young's modulus of the alginate scaffold decreased over time [25]. It was found to be closer to that of the brain cancer tissue at around 7 kPa [54] than cell cultures on polystyrene plates (2-4 GPa) [55]. Tang et al. had reported that the stiffness of the GSC-encapsulated tumour core was 2.8 ± 0.6 kPa, whereas the less dense peripheral region was 0.9 ± 0.2 kPa (consisting of encapsulated neural progenitor cells and astrocytes) [31]. The stiffness of the peripheral region was meant to be similar to that of healthy brain tissue, which is claimed to be 1 kPa [44]. In another study conducted by Tang et al. two types of mechanical properties or stiffnesses were created to represent glioblastoma and healthy brain tissue in the ECM regions, respectively [28]. A bioprinted region with a stiffness of 21 kPa was referred to as the stiff model, while another bioprinted region with 2 kPa was referred to as the soft model, considering that the matrix stiffness in glioblastoma tissues could rise to 26 kPa from the previous clinical investigations [56][57][58][59]. The tumour cells and epithelial cells showed high viability in their respective hydrogel environment after one week of being cultured. Therefore, Tang et al. concluded that the stiffness-patterned models may be ideal for mimicking different stages of glioblastoma development because the tumour and endothelium regions were intended to have stiffness simulating their native states and both invasion patterns have previously been seen for glioblastoma cells [28].

Biocompatibility and Cellular Response
The biocompatibility of bioinks was thoroughly explored. When considering a possible material for medicinal usage, cytotoxicity should be considered. The live/dead cell assay was used in most of the studies to ensure cell-to-material chemical contact did not cause cytotoxicity. Multiple diverse populations of malignant and supportive stromal cells make up the brain tumours, and these intricate cellular interactions are critical for tumour survival, growth and progression. Intratumoural cell heterogeneity in glioblastoma is extensive, with contributions from astrocytes, neurons, macrophage/microglia and vascular components. The researchers also showed the advantage of using 3D bioprinted tumour models by bioprinting several different cell types simultaneously [17,22,27,28,30,31]. Bioprinted HUVECs and LFs in GAF hydrogel were cultured until blood vessels with lumens developed. The multicellular tumour spheroids were then seeded into the blood vessel layer and incubated until the blood vessel layer's endothelial cells migrated into the multicellular tumour spheroids and displayed angiogenesis, while some cancer cells penetrated the blood vessel layer [17]. In another study, migration of HUVECs towards the glioblastoma cells was seen in 3D co-culture models. The migrating HUVECs in the stiff model had a sprouting blood vessel shape and were in close contact with the glioblastoma cells, whereas the HUVECs in the soft model had an expansive-growth morphology with no apparent sprouting. The expression of the angiogenic marker vascular endothelial growth factor (VEGF) in the tumour cells in the co-cultures of both the soft and the stiff hydrogel was significantly increased compared to the tumour-only models [28].
Furthermore, Heinrich et al. found that the tumour cells migrated towards macrophages as opposed to towards tumour cells themselves or to an empty well [19]. There was significant upregulation of glioblastoma markers in the co-cultured model with RAW264.7 macrophages. Therefore, the research demonstrated that tumour cells may recruit macrophages to their site and train them to maintain or enhance tumour survival and growth [19]. In another study conducted by Tang et al. the 3D tetra-culture model (macrophages were combined with GSCs within the tumour core surrounded by astrocytes and neural progenitor cells) showed an elevation of the glioblastoma tissue-specific gene set when compared to a GSC spherical culture [31]. The presence of macrophages enhances genes that would promote hypoxia and pro-invasive transcriptional profiles, demonstrating that the tetra-culture model or multicellular culture system recapitulates the transcriptional states seen in patient-derived glioblastoma tissues [31].

Drug Response and Clinical Value
The included studies found that 3D-cultivated models had higher cell viability and higher IC 50 than 2D-cultured models. Wang et al. hypothesised that the combination of GSC proliferation, a hypoxic environment and the activation of epithelial-mesenchymal transition (EMT) resulted in increased drug resistance in 3D-cultured cells [15]. Cancer stem cells (CSCs) are not only responsible for chemoresistance [60] but also tumourigenicity. To evaluate the tumourigenicity of CSCs in nude mice, 1 × 10 4 cells were harvested from 2D or 3D conditions, and 3D-cultured cells were shown to be more tumourigenic than 2D-cultured cells, indicating that the stemness qualities of glioma cells were improved in 3D bioprinted scaffolds [15].
Tang et al. reported that the stiff condition and co-culture with endothelial cells enhanced glioblastoma drug resistance as compared to a spherical culture control. Furthermore, the stiff co-culture model showed the highest tumour cell viability after TMZ treatment, whereas the soft model showed higher TMZ sensitivity, indicating that the cancer drug, TMZ, induced cell cycle arrest and halted cell division, thus inhibiting cell proliferation [28].
Shell-GSC23/core-U118 hydrogel microfibers showed higher resistance to TMZ with a significantly lower methylation rate of O 6 -methylguanine-DNA methyltransferase (MGMT) promoter when compared to core-U118 hydrogel microfibers [24]. Recent research has proven that MGMT is a key component in tumour prognosis [61,62]. Methylation of the MGMT promoter can silence the gene in cancer cells and limit their ability to repair DNA, rendering cancer cells more vulnerable to TMZ. The higher the degree of MGMT methylation, the lower the MGMT protein expression, resulting in a favourable prognosis in the setting of TMZ administration [24].
Various drugs and combinations have been used to evaluate treatment resistance and to identify a patient-specific drug combination [22]. The ataxia-telangiectasia mutant kinase, which activates critical proteins that initiate DNA-damage-response pathways, was inhibited by combining CIS with KU [63]. Surprisingly, CIS + KU reduced the survival cell percentage of glioblastoma-28-on-a-chips; however, glioblastoma-37-on-a-chips were less sensitive to the same treatment. When compared with glioblastoma-37-on-a-chip, glioblastoma-28-on-a-chip was more susceptible to O 6 BG (a pseudosubstrate of MGMT) and MX (a base excision repair pathway inhibitor), including the combination of O 6 BG + MX and the combination of CIS + KU + O 6 BG + radiation. The resulting ex vivo glioblastoma model can be used for the identification of an optimal treatment for patients with the aid of personal bioinformatics analysis [22].
Tumour cells in 2D and 3D models may respond differently to the same treatment. Furthermore, drug test findings acquired from animal models cannot fully reflect what would be observed in the human body due to cross-species variations, as more than 95% of drugs that are successful in animals are not as effective in humans [3]. In the end, the majority of drugs would fail in the pre-clinical testing. Hence, a 3D bioprinted tumour model, which gives a more accurate representation of the tumour microenvironment, is an ideal tool for evaluating drug efficacy.

Study Limitations
There are some limitations to this systematic review. First, no specific checklist has been developed for the analysis of the risk of bias in in vitro studies. Thus, we used the JBI checklist to assess the studies that were included. Moreover, studies published in languages other than English were excluded. Unpublished reports or studies published in languages other than English may be missed unintentionally. Furthermore, while the use of 3D bioprinting for cancer modelling has been widely researched in vitro, only a few studies have been conducted in vivo. This limits our understanding of the clinical value of 3D bioprinting in in vivo conditions. Furthermore, there was heterogeneity among the studies, as the duration of observation and measurements as well as the cell lines used varied across studies, which made it unsuitable to quantitatively combine the findings.

Conclusions
In this systematic review, both in vitro and in vivo studies demonstrated that bioinks have a strong potential to imitate the 3D microenvironment of the native brain tumour tissue while promoting cell viability and proliferation. The difference between the effects of therapeutic drugs on 2D and 3D cultures was reported. Monolayer or 2D cell culture models fail to capture the heterogeneity and complexity of the tumour microenvironment. However, 3D bioprinted models allow the incorporation of more than one cell type, simulate the 3D geometry of native tumour tissues, and supply accurate nutrition and oxygen gradients. Some studies reported an increase in the drug efficacy in 3D models with a printed porous structure or a vascular network because more regions or higher surface volumes of the tumour were subjected to the drug. In drug screening, a combination drug of TMZ, BCNU, doxorubicin (DXR) and cordycepin (COR) was reported to have higher IC 50 when tested with the 3D bioprinted glioblastoma model than the 2D cell culture model. Furthermore, multi-drug treatments (e.g., TMZ + SU, TMZ + BEZ235, and niraparib (NIRA) + BEZ235) have shown a greater therapeutic response than single-drug treatments.
Overall, the recent developments on in vitro 3D bioprinted glioblastoma models presented in this paper have contributed to a better understanding of the fabrication techniques of bioinks, characteristics and effectiveness of bioinks, tissue bioprinting strategies and the discovery of potential treatment targets. In several studies, the combination of bioinks of different materials, including natural and synthetic materials, was explored to construct a desired structure that supported the tumour microenvironment. We suggest that future research should focus on bioprinting of co-cultured tumour models with a vascular component, astrocytes, and microglia/monocyte/macrophages to represent a more precise heterogenous tumour microenvironment. Since most recent studies still use commercial cell lines, primary cells and cancer stem cells derived from patients can also contribute to a more precise model and effective evaluation. Additionally, the physical parameters of bioinks, such as stiffness and porosity of the scaffolds, must be considered in order to model the true state of the ECM. A sophisticated 3D glioblastoma tumour model is essential for high-throughput drug screening to replace conventional animal trials and may drive the realisation of personalised treatment in the future.